[codex] Add benchmark leakage audit assistant

[Davifek]

@algora-pbc /claim #16

Summary

Adds a distinct benchmark-leakage-audit-assistant/ slice for the AI-Powered Research Assistant Suite bounty. This module focuses on ML/research evaluation hygiene before publication or release review.

It detects:

• train/test overlap by record ID or normalized content fingerprint
• benchmark contamination in the training corpus
• final holdout or test-set use during model selection
• weak split provenance such as missing seed or manifest hash
• missing reproducibility packet artifacts

Demo

Included short demo artifact: benchmark-leakage-audit-assistant/demo.gif

Run locally:

node benchmark-leakage-audit-assistant/demo.js

Verification

Passed locally:

node benchmark-leakage-audit-assistant/test.js
node benchmark-leakage-audit-assistant/demo.js
git diff --cached --check

The tests cover a contaminated project that is blocked, a clean project that passes without false positives, and reviewer-ready evidence/remediation fields for every finding.

Notes

• Dependency-free Node.js standard library implementation
• Synthetic sample data only
• No credentials, external services, or network access required
• Includes a requirement map for AI-Powered Research Assistant Suite #16