Port PerformanceStudio's wait-row layout: an Auto benefit column that never clips

[erikdarlingdata]

Fixes #4576. Part of #4511.

Why

PerformanceStudio (erikdarlingdata/PerformanceStudio@78a3370, refined @80de6fc) fixed the Wait Stats card's rows so the flexible column is the duration text, not the bar; the wait-type name is capped at 150px with an ellipsis and a tooltip; and the trailing "up to N%" benefit sits in an Auto column so it is never the thing that gets clipped. Every row also carries a tooltip now.

PerformanceMonitor's wait grid (PlanViewerControl.Properties.cs's ShowWaitStats, shared by the Darling Viewer and Lite) had none of this: a fixed name column sized off the longest wait-type string, a fixed 300px bar column, and no benefit column at all — the benefit percentage the analyzer already scores per wait type (BenefitScorer.EmitWaitStatWarnings) was never surfaced next to the row it belongs to.

What changes

• ShowWaitStats's grid columns are now *, Auto, *, Auto (name, bar, duration, benefit) instead of three fixed-width columns. The name and duration are the two star columns that ellipsize (TextTrimming.CharacterEllipsis) under pressure; the bar (60px) and benefit are Auto so neither is ever clipped.
• The wait-type name is capped at 150px (MaxWidth) with an ellipsis and a "{type} — {category} wait" tooltip.
• The duration text gets a "{ms} ms across {n} waits" tooltip.
• The header's total-ms text now also carries a tooltip ("{total} ms of waits across {n} wait types"), cleared when a statement has no waits so a hover doesn't answer with a stale count.
• New: a benefit column reading "up to N%" for any wait type the analyzer scored (BenefitScorer.EmitWaitStatWarnings's "Wait: {type}" finding), with a "Up to N% of this statement's runtime could be recovered by removing {type} waits" tooltip. Absent when there's no matching finding, no score, or the score is zero.
• New pure helper: PerformanceMonitor.PlanAnalysis.WaitRowText.Benefit(waitType, statementWarnings) does the join and formatting (case-insensitive match on "Wait: {type}", reusing PlanWarningDisplay.FormatBenefitPercent's rounding, now internal). The WPF code-behind calls it; nothing else changed in the join.
• ShowWaitStats's signature grew a List<PlanWarning> parameter (the statement's PlanWarnings); the one call site (PlanViewerControl.Rendering.cs) was updated.

Lite: gets this too (shared control)

Lite hosts the same PerformanceMonitor.Ui.PlanViewerControl, so this lands there with no separate change.

Not ported / PM-ahead

• Wait-category colors stay as-is; that's Wait-category colours: PerformanceStudio's contrast-checked palette #4578 (already merged as Wait-category colours: a laid-out, contrast-checked palette #4583).
• The Wait Stats header text's own benefit suffix is Plan viewer: make the wait stats card collapsible, closed by default #4571's scope, not touched here.
• This only touches the row layout inside ShowWaitStats.

Test plan

New pin file: Darling/Darling.Tests/ViewerWaitRowBenefitTests.cs — pins WaitRowText.Benefit: a positive match, case-insensitive match, the shared 100%-and-above whole-number format, no matching finding, a null score, and a zero/negative score all resolving to null.

• RED on dev (compile-only, all-new helper): copying the test file into a detached worktree at origin/dev (49c5b6a) fails with 7× CS0103: The name 'WaitRowText' does not exist in the current context — the type doesn't exist there yet.
• Mutation RED (runtime, on the fix branch): changed the lookup key from "Wait: " + waitType to "Wait " + waitType (dropping the colon) — 3 of 7 tests failed (Benefit_ReturnsUpToNPercent_WhenAMatchingFindingHasAPositiveScore, Benefit_MatchesCaseInsensitively_OnTheWaitType, Benefit_UsesTheSharedWholeNumberFormat_AtAndAbove100). Reverted; rebuilt; all 7 green again.
• GREEN, -class Darling.Tests.ViewerWaitRowBenefitTests: Total: 7, Errors: 0, Failed: 0, Skipped: 0.
• GREEN, the full required class list (ViewerWaitRowBenefitTests + ShowPlanParserCondAndMultiplePlanTests, ActualPlanRequestTests, ActualPlanDispatchTests, ActualPlanResultParseTests, QueryModificationDetectorTests, ActualPlanCaptureLoopTests, ActualPlanGatingTests, ReproScriptBuilderHardeningTests, DarlingAnalysisPipelineTests, DarlingMcpPlanToolsSurfaceAndSqlTests, DarlingMcpPlanToolsLivePostgresTests, McpPlanAnalysisEnvelopeTests, SerialLoopStoreSizeSourceTests, TsqlConventionGuardTests, DocCommentHygieneTests): Total: 238, Errors: 0, Failed: 0, Skipped: 2 (the two live-Postgres tests, gated on DARLING_TEST_PG, skip on this rig as expected).
• Builds, all -p:EnableWindowsTargeting=true, analyzer-warnings grep empty on every log: Darling.Tests (Release), PerformanceMonitor.Ui, Lite/PerformanceMonitorLite.csproj, Lite.Tests, Darling/PerformanceMonitor.Darling.Viewer — all 0 warnings / 0 errors.
• PerformanceMonitor.PlanAnalysis itself builds 0/0 (the internal visibility change to FormatBenefitPercent doesn't break anything outside the assembly; it's still only called from within PerformanceMonitor.PlanAnalysis).

Screenshot plan (pure layout, can't assert visually from here)

Open a captured plan with wait stats and at least one wait type the analyzer scored a benefit for (e.g. one with meaningful CXPACKET or PAGEIOLATCH_SH time), in both the Darling Viewer and Lite (shared control):

1. Narrow the window/panel to roughly its minimum width. The wait-type name and duration text should each shrink and end in an ellipsis rather than either wrapping, clipping mid-word, or growing a horizontal scrollbar (there is none — HorizontalScrollBarVisibility="Disabled" on WaitStatsHeader's ScrollViewer was already in place).
2. The "up to N%" benefit text at the row's right edge should be fully visible at every width, never truncated.
3. Hover the wait-type name: tooltip reads "{type} — {category} wait". Hover the duration: tooltip reads "{ms} ms across {n} waits". Hover the benefit text: tooltip reads "Up to N% of this statement's runtime could be recovered by removing {type} waits". Hover the "Wait Stats — Nms total" header: tooltip reads "{N} ms of waits across {n} wait types".
4. A wait type the analyzer didn't score (or scored at 0%) shows no benefit text in that row — no blank cell artifact, no "up to 0%".
5. Colors are unchanged from before this PR (that's Wait-category colours: PerformanceStudio's contrast-checked palette #4578/Wait-category colours: a laid-out, contrast-checked palette #4583's territory).

Follow-up: the benefit text is now a whole percent (a13ed6e)

PerformanceStudio's desktop plan viewer formats the wait-row benefit as $"up to {benefitPct:N0}%" — a whole number, no decimal. WaitRowText.Benefit was formatting with PlanWarningDisplay.FormatBenefitPercent, which keeps one decimal below 100%. Fixed: WaitRowText.Benefit now formats with N0 directly, matching PerformanceStudio's rounding at every value, not just at and above 100%. FormatBenefitPercent goes back to private (it's used only by the plan-warning header's own display, a different format).

Pins updated in ViewerWaitRowBenefitTests.cs: 42.6 → "up to 43%", 12.0 → "up to 12%", 100.0 → "up to 100%" (unchanged), and a new case, 0.4 → "up to 0%" (PerformanceStudio still shows the row when the score is positive but rounds to zero, since the > 0 gating condition is on the raw score, not the rounded text).

RED on the pre-fix commit (7fdbd52): copying the updated test file into a detached worktree at that commit and running -class Darling.Tests.ViewerWaitRowBenefitTests fails 3 of 8: Benefit_ReturnsUpToNPercent_WhenAMatchingFindingHasAPositiveScore ("up to 43%" expected, "up to 42.6%" actual), Benefit_MatchesCaseInsensitively_OnTheWaitType" ("up to 12%" vs "up to 12.0%"), and Benefit_RoundsAFractionalScore_ToTheNearestWholePercent` ("up to 0%" vs "up to 0.4%").

GREEN on the fix: -class Darling.Tests.ViewerWaitRowBenefitTests → Total: 8, Errors: 0, Failed: 0, Skipped: 0.

Census family, run on this rig: CommentFilterAdoptionTests, DocCommentHygieneTests, and every *Census*Tests class that doesn't need a local Postgres server (EntraProviderPackageCensusTests, FileGrowthRiseUnitCensusTests, McpPayloadContractCensusTests, McpServiceParameterDiSeatCensusTests, PgSettingScrubCandidateCensusTests, SameStatementPileupSourceCensusTests, WebExceptionTextCensusTests) all pass alongside the updated pin: Total: 285, Errors: 0, Failed: 38, Skipped: 0 where every failure traces to LivePostgresStoreFixture.InitializeAsync (no Postgres server on this rig) — a pre-existing environment gap, not a regression from this change. DailyDeadlockWindowCensusTests, MeasurementContractCensusTests, PlanForceActionDetailCensusTests, RemoteCollectorServiceCancellationCensusTests, StoreApplicationNameCensusTests, and StoreSessionTimeZonePinCensusTests weren't run separately for the same reason (they also need a live Postgres fixture); CI decides them.

Builds, -p:EnableWindowsTargeting=true, analyzer-warnings grep empty: Darling.Tests (Release) and PerformanceMonitor.Ui — both 0 warnings / 0 errors.

CHANGELOG

SECTION: Changed
ENTRY: - The plan viewer's Wait Stats rows show each wait's potential benefit and never clip ([#4595]) - As in PerformanceStudio's desktop viewer, each wait row gains an "up to N%" benefit column taken from the plan's matching wait finding, the name and duration columns take the spare width with an ellipsis and a tooltip, and the bar and benefit columns size to their content so the percentage is never cut off.
REF: [#4595]: #4595