Deterministic review-effort / complexity score + review-time ROI signal

[JSONbored]

Phase 1 — Review output, noise & trust

Context

We already compute added-line counts and file priority (src/review/review-diff.ts: addedLineCount, diffFilePriority) and a size gate, but never surface a review-effort estimate. src/review/public-stats.ts currently credits a flat 20 minutes saved per reviewed PR — not credible.

Goal

A deterministic per-PR review-effort/complexity score, surfaced in the review comment and used to compute a credible review-time ROI figure.

Requirements

• Pure deterministic estimator (files touched, added lines, file-type mix via path-matchers, hunk spread) → a 1–5 complexity band + minutes estimate. No AI; a review.effort_score toggle controls display.
• Replace the flat MINUTES_SAVED_PER_PR in src/review/public-stats.ts with the per-PR estimate; surface the aggregate only in the self-host maintainer dashboard/stats feed (src/review/stats.ts).

Deliverables

• Estimator module + tests.
• A compact "review effort: N/5 (~M min)" line in the unified comment (src/review/unified-comment.ts), display gated by config.
• public-stats.ts + stats.ts wired to the per-PR estimate.

Expected outcome

• Credible, grounded review-time-saved reporting; a quick complexity read for maintainers triaging the queue.

Effort

S.

[linear-code]

JSO-2035