Merge train 4843-19 (CI only; not for merge)

[erikdarlingdata]

Merge train, not for merge. CI runs once on dev plus the squash of each PR below, in this order. When it is green, each is squash-merged by itself in the same order, dev's tree is checked against this branch's, and this PR is closed unmerged.

Part of #4966, #5097 and #5158.

PRs in this train, in order

• MCP: Darling's blocking stats and PostgreSQL deadlock and log event tools say where their data starts (part of Say where the data starts on every time-ranged tab, in Lite and both Darling viewers #4966) (MCP: Darling's blocking stats and PostgreSQL deadlock and log event tools say where their data starts (part of #4966) #5205)
• The store accepts a known plan digest for a row and keeps that plan alive (inert, query_stats renders and ships the same ~200 plans every minute: 47x more plan XML than is distinct, and the reason its slow runs are slow #5158) (The store accepts a known plan digest for a row and keeps that plan alive (inert, #5158) #5174)
• Add a --diagnostics-bundle verb that writes one aliased, attachable diagnostics file (Slow reads and slow collector runs can't be diagnosed from the product: record each slow call's arguments, fix the stall probe's waits, keep store statement history, and export a diagnostics bundle #5097) (Add a --diagnostics-bundle verb that writes one aliased, attachable diagnostics file (#5097) #5218)

Train head 9de997ed4, tree f33875cb2, base dev 7aca8830a. No member moves the tools/list budget; the next train carries one that does, alone.

A fourth pull request was excluded deliberately. It builds on the second member here and still carries that member's OLD commits, so its copy of one live test file differs from the version being landed. Each member lands as its own squash, so its squash would have quietly reverted the second member's fix, and the test would then fail on a later train looking like a fresh flake rather than a reversion. Its author is re-merging it against dev after this train lands and verifying that file byte-for-byte.

Local merged-tree checks: 1,803 tests, 0 build warnings; the store-upgrade class at 9 and the hygiene class, both matching dev. The repository-source scan over 403 classes reports no new failure. The live-store hygiene guard was emulated against this merged tree and reports 0 offenders — that guard cannot execute on macOS, so until a few hours ago a violation in it was visible only in CI, twenty minutes into a run.

[erikdarlingdata]

Superseded by train 19b. One member's own aliasing-hardening test fails in CI with the refusal message null - nothing refused - and that test is absent on dev, so it is new and attributable. The other two members ride 19b. Closing unmerged.