Add ReviewSLA step 1: first-response clocks and report mode

[willson556]

Step 1 of Proposal 2's rollout (first-response SLA, rollout plan): the clock code and report mode. Nothing is scheduled, so nothing changes for authors or reviewers. The scheduled job, with reminders and the Confluence dashboard, comes in step 2's PR.

• ReviewSLA/clock.js rebuilds a PR's first-response clock from its ready and draft events and its reviews. It keeps two measures:
  • RFC: the RFC baseline's method, in wall-clock hours, from the first ready event to the first review by anyone but the author.
  • SLA: the plan's clock rules, in business hours against a 4 h or 7 h target. The clock restarts each time the PR is marked ready again. Only an approval, changes requested, or a comment review with a body or inline comments stops it.
• ReviewSLA/report.js is report mode. Like CIMetrics, it runs locally with gh, and it writes weekly numbers per repo plus a per-PR TSV for spot checks. Events after --until are ignored, so a past range reads the same whenever it's run.
• ReviewConfig/hours.js adds business hours (10:00 to 17:00 Pacific, minus weekends and holidays.json), next to the PTO calendar code it builds on. pto.js now exports its date helpers.
• The test workflow is renamed to "Review action tests" and also runs on changes to ReviewSLA/.

Step 1's check: over Jun 26 to Sep 24, does the RFC measure reproduce the RFC?

	RFC	This PR, the RFC's PRs	This PR, every PR ready in range
StaflLib	64 h median, 232 h p90, 21% within 8 h	139 PRs: 64.5, 232.3, 21%	173 PRs: 72.0, 198.7, 19%
coit-tower	5 h, 30 h, 70%	71 PRs: 5.4, 29.5, 70%	same

For the PRs the RFC measured, every timing matches its source data to within 0.1 h. The full run times 34 more StaflLib PRs. The RFC sampled StaflLib's 300 most recent PRs by creation date, and these were opened before Jun 26 but marked ready during the range.

The business-hours baseline for the gates, over the same range:

	First responses timed	Median	p90	Within target	Unreviewed past target at Sep 24
StaflLib	171	14.4 bh	43.3 bh	34%	8
coit-tower	72	4.1 bh	11.4 bh	64%	4

Testing:

• node --test AssignReviewers/ ReviewConfig/ ReviewSLA/: 55 cases, 21 of them new. They cover the plan's clock cases that don't need stacks or assignees: 16:30 Friday to 13:30 Monday, a holiday, which reviews count, draft and ready again, a PR reviewed as a draft, unreviewed PRs before and after their target, a PR closed unreviewed, events after the range, the switch to standard time, and bot and merge-queue PRs.
• Ran report mode against StaflLib and coit-tower, with the results above.

🤖 Generated with Claude Code

[willson556]

• Pick reviewers more fairly: own PRs don't count as load, ties go to the longest wait #45
• Add the scheduled ReviewSLA job: comments, reminders, digest, dashboard #44
• Add ReviewSLA step 1: first-response clocks and report mode #43 👈 (View in Graphite)
• AssignReviewers: skip people who are out, from the Rippling PTO calendar #41
• Assign reviewers on this repo's own PRs #42
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[staflsystemsci]

Review SLA

Nobody owes a review on this PR right now.

_{Due times are in business hours, 10:00 to 17:00 Pacific: 4 h for a PR under 250 added lines, 7 h otherwise, and 4 h for a re-review. Mid-stack, an assignee owes a response on the lowest PR they haven't approved. Dashboard}

[willson556]

Merge activity

• Oct 2, 10:56 PM UTC: A user started a stack merge that includes this pull request via Graphite.
• Oct 2, 10:58 PM UTC: Graphite rebased this pull request as part of a merge.
• Oct 2, 10:58 PM UTC: @willson556 merged this pull request with Graphite.