docs(cookbooks): add Daytona RL cookbook

[shfunc]

Note

Low Risk
New cookbook and benchmark scripts only; no changes to core HUD libraries or production services.

Overview
Adds cookbooks/daytona-rl, an end-to-end example that runs GRPO training on Daytona sandboxes using the same HUD env.py for eval and rollouts (pytest-graded “fix calc.py” task with 24 deterministic bug variants).

The cookbook wires up DaytonaRuntime + a Dockerfile.hud image, content-addressed snapshot names (snapshot.py), and warm-pool helpers (pool.py) sized to match batch concurrency. train.py runs the documented 10-step / 128-wide run (chunked forward_backward to avoid 413s, resume via train_run.json, pool teardown on exit). bench/ adds reproducibility scripts (baseline pass rate, concurrency ladder, warm-pool A/B, cleanup) plus committed train_run.json / train_curve.svg for the published before/after numbers.

^{Reviewed by Cursor Bugbot for commit 1c8ad44. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit bfb3805. Configure here.}