feat: add lab result submission contract

[0xnavarro]

Summary

Makes ReflexBench easier to integrate from model-lab and evaluation-platform CI.

• adds FOR_LABS.md
• adds a versioned JSON Schema for result-submission envelopes
• adds a canonical template
• adds a dependency-free validator CLI
• adds unit tests for validation and disclosure requirements
• links the lab integration path from the README

Why

A benchmark becomes easier to adopt when third parties can produce machine-readable evidence directly from CI instead of relying on screenshots or bespoke issue text. This keeps raw model results, model+harness results, corpus identity, engine identity and development-contamination disclosure explicit.

QA

• 50/50 unit tests PASS
• python3 tools/validate_submission.py templates/result-submission-v1.json PASS
• python3 tools/check_public_release.py . PASS
• python3 tools/check_claims.py . PASS
• git diff --check PASS

No benchmark tasks, scoring, v1 corpus identity or existing result values change.