Add challenge evaluator calibration guard

[zergzorg]

Summary

• Adds a self-contained challenge-evaluator-calibration-guard module for Scientific Bounty System issue Scientific Bounty System #18.
• Validates synthetic seed coverage, evaluator role/conflict exclusions, severity drift, criterion drift, inter-rater spread, and score-normalization anchors before live scoring opens.
• Emits deterministic JSON/Markdown/SVG outputs plus a short MP4 demo artifact for reviewer verification.

Validation

• npm run check
• npm test
• npm run demo
• ffprobe -v error -show_entries format=duration,size -show_entries stream=codec_name,width,height -of default=noprint_wrappers=1 challenge-evaluator-calibration-guard/reports/demo.mp4 -> H.264, 960x540, 5s
• git diff --check
• rg -n "(password|secret|wallet|paypal|bank|passport|tax|kyc|private key|api key|token)" challenge-evaluator-calibration-guard || true -> no matches

Non-overlap

This is a focused calibration seed readiness guard before live sponsor/reviewer scoring. It avoids the existing intake, rubric-readiness, scoring/arbitration, reviewer consensus, review-integrity, appeal, cancellation/no-award, benchmark leakage, workspace privacy, data-room access, license/dependency, human-subjects, withdrawal, and embargo-release slices.

Synthetic data only; no credentials, private identity data, payout details, or private dashboard data are included.

@algora-pbc /claim #18