fix(review): count owner reopen-then-merge as a real reversal signal

[JSONbored]

Closes #7985.

Summary

• recordReversalSignals previously excluded EVERY repo-owner reopen of a bot-closed PR from the reversal signal (treated as "administrative re-queue, not a dispute"). That blanket exclusion is right for a bare reopen — genuinely ambiguous on its own — but wrong once the owner actually merges the PR afterward: that's an unambiguous "the bot's close was wrong" correction, not routine administration.
• This exact pattern (bot closes, owner reopens, owner merges within minutes) was every single one of the 2026-07-21/22 metagraphed false-positive incidents (fix(review): paginate preview-url.ts's PR-comment and check-run GitHub reads #7469, Epic: dual-path repo provisioning — BYOR + APR #7589, Design spec: idea-completion determination + customer-facing status for APR (no GitHub UI exposure) #7591, fix(ui): give four app-panel form controls a programmatic accessible name #7594) — and because none of them ever got recorded as a reversal, the accuracy/self-correction system had no signal that anything was wrong on the day it mattered most, part of the reason the homepage "Decision" accuracy metric looked misleadingly close to perfect.
• Fix: an owner reopen of a bot-closed PR now writes a time-bounded owner_reopen_pending_reversal marker (via recordAuditEvent) instead of being silently dropped. If a merge for that same PR follows within OWNER_REOPEN_MERGE_WINDOW_MS (6h), it's promoted to a real reversal_reopened event in both review_audit (what reversalRate/calibration read) and audit_events. A bare reopen with no follow-up merge, or a merge outside the window, still records nothing extra — same "genuine correction, not noise" bar a contributor reopen already has to clear. A bot reopening itself is still never a human-disagreement signal.

Test plan

• npm run typecheck
• New tests in test/unit/outcomes-wire.test.ts (describe("owner reopen + merge within the window (#7985)")): promotes reopen+merge to a real reversal; a plain merge with no marker records nothing; a stale marker outside the window records nothing; an owner reopen when the last bot action wasn't a close writes no marker
• Full unsharded npm run test:coverage: 1088/1088 files, 20312 tests, 0 failures

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[codecov]

Codecov Report

❌ Patch coverage is 78.94737% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 89.67%. Comparing base (a3dc73e) to head (cffa2f7).
⚠️ Report is 6 commits behind head on main.

Files with missing lines	Patch %	Lines
src/review/outcomes-wire.ts	78.94%	2 Missing and 2 partials ⚠️

❌ Your patch status has failed because the patch coverage (78.94%) is below the target coverage (99.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8074      +/-   ##
==========================================
- Coverage   91.83%   89.67%   -2.16%     
==========================================
  Files         734        1     -733     
  Lines       75086      184   -74902     
  Branches    23068       50   -23018     
==========================================
- Hits        68952      165   -68787     
+ Misses       5034       12    -5022     
+ Partials     1100        7    -1093     

Flag	Coverage Δ
rees	?
shard-1	26.63% <36.84%> (-28.71%)	⬇️
shard-2	89.13% <78.94%> (+33.34%)	⬆️
shard-3	26.63% <36.84%> (-23.94%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/review/outcomes-wire.ts	89.67% <78.94%> (-1.29%)	⬇️

... and 733 files with indirect coverage changes