Docs-Sync #402: two README activity-count claims sit outside DocumentationCountsTests #404

Source PR: #402 (Fix typos and tighten wording in the root Markdown files)

Affected surface: tests/NodePilot.Mcp.Tests/DocumentationCountsTests.cs (guard coverage for README.md)

#402 was a copy-editing pass whose commit message states "the phrases and counts the documentation
guards read are left intact". That held for every guarded phrase — all 18 rows still match the code —
but the pass silently rewrote two unguarded claims from 27 activities to 29:

• README.md reference table: [All 29 activities]
• README.md solution tree: NodePilot.Engine/ WorkflowEngine, 29 activities

while the guarded highlights bullet (with 27 activity types) stayed at 27, leaving the README
contradicting itself. The wrong numbers are corrected in #403; this issue is about the gap that let
them drift, which is test code rather than documentation and so was deliberately not auto-applied.

DocumentationCountsTests pins exactly one of the README's three activity-count claims:

yield return Row("README.md", @"with (\d+) activity types", activities, "activity types (README highlights)");

The file's own comment records that the "Beyond the N executable Activity types…" row was retired
because "the count is still guarded twice — in the highlights row above, and in both language
versions of the doc site's concepts/workflows page". That reasoning covers the doc site, but it
leaves the README's other two mentions unguarded, and #402 is the first case of them actually
drifting.

Un-applied findings

• DocumentationCountsTests has no row for the README reference-table claim
  ([All (\d+) activities]). Adding one is a test change, not a doc fix, and whether to guard a
  link label is a maintainer decision — needs a human decision.
• DocumentationCountsTests has no row for the README solution-tree claim
  (WorkflowEngine, (\d+) activities). Same reasoning; it also sits inside a fenced block, so a
  guard row would need a pattern that tolerates that context — needs a human decision.
• Worth deciding at the same time whether the retired-row rationale quoted above should be
  revised, since it is the reason these two mentions were left uncovered — ambiguous, and it is
  a comment about test policy rather than a factual drift.

This is an automated docs-drift finding and needs triage. The counts themselves are already correct
on the #403 branch; nothing here blocks that PR.