[codex] Add Modal B200 submission rate limit

[msaroufim]

What changed

Add a Modal B200 submission cap of 1 per hour and enforce it before submissions are created or queued.

Why

The existing rate-limit system is keyed by leaderboard and mode category. It does not model GPU-specific limits, so it could not express a Modal B200 cap.

Impact

Modal B200 submissions are now rejected once a user has already submitted to that GPU within the last hour. The check applies to both immediate submissions and queued async API submissions.

A migration adds requested_gpus to leaderboard.submission so the limiter can make that decision before any run rows exist.

Validation

• uv run pytest tests/test_submission.py -q
• uv run python -m compileall src/libkernelbot src/kernelbot/api

Notes

I added DB-backed tests for the new GPU rate-limit path, but I could not run them locally because Docker was not running for the Postgres test container.

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
src/libkernelbot
backend.py
consts.py
leaderboard_db.py					290
submission.py					52, 54, 69
utils.py
Project Total

This report was generated by python-coverage-comment-action