Add review calibration bench

[FeeeeelixWong]

/claim #15

Summary

• add a self-contained peer-review calibration bench for the Community & User Reputation System bounty
• compare reviewer scores against consensus panels to detect leniency/severity drift, reproducibility blind spots, and calibration gaps
• produce role/status labels, coaching actions, quarantine recommendations, digestable evidence, and a small demo video

Why this is scoped to issue #15

This focuses on the review-quality side of community reputation: calibrated reviewers, trust signals, coaching tasks, and reproducible evidence for reputation changes. It avoids the broader account reputation/appeals work already covered by other PRs.

Demo

• review-calibration-bench/docs/demo.mp4
• review-calibration-bench/docs/demo.svg

Verification

• npm run check
• npm test
• npm run demo
• ffprobe review-calibration-bench/docs/demo.mp4
• git diff --check