🐛 Align RL rewards and terminal outcomes

[flowerthrower]

🤖 AI text below 🤖

Description

A compilation that exhausts its pass budget now receives the same final reward as explicit termination: the circuit score once if synthesis, layout, and routing are complete, otherwise the existing no_effect_penalty. This terminal reward replaces the last pass's intermediate reward; earlier improvement rewards and no-effect penalties remain unchanged. Inference raises for incomplete terminal results.

Hard-budget endings, pass failures, and pass timeouts terminate without PPO value bootstrapping. External truncations retain their reward and bootstrap from the final observation. Both flat and GNN observations include the normalized remaining pass budget; GNN global features increase from 36 to 37. Models must be retrained. The upgrade guide documents the observation and error-info changes.

Depends on #800 and remains the final draft in the paper stack:
#831 → #755 → #758 → #794 → #795 → #796 → #797 → #798 → #799 → #800 → #830.
Only #831 targets main. Keeping this policy in #830 lets its tests exercise #799's rewards and #800's GNN together.

Validation: all 117 non-training compilation tests and full lint pass, including all 68 RL/GNN tests. Coverage includes complete/incomplete horizons, explicit termination, failure/timeout penalties, unchanged intermediate rewards, remaining-budget observations, traces, flat/GNN PPO bootstrap behavior, and saved GNN inference. Fresh CI remains pending.

AI assistance: Codex implemented the policy, extended the existing tests, and updated the changelog, upgrade guide, and PR description.

Checklist

• The pull request only contains commits that are focused and relevant to this change.
• I have added appropriate tests that cover the new/changed functionality.
• I have updated the documentation to reflect these changes.
• I have added entries to the changelog for any noteworthy additions, changes, fixes, or removals.
• I have added migration instructions to the upgrade guide (if needed).
• The changes follow the project's style guidelines and introduce no new warnings.
• The changes are fully tested and pass the CI checks.
• I have reviewed my own code changes.

If PR contains AI-assisted content:

• Any agent that created, edited, or submitted GitHub content was explicitly authorized for that scope, as required by our AI Usage Guidelines.
• Every agent-authored or agent-edited public text body begins with the visible disclosure 🤖 *AI text below* 🤖 (titles are exempt).
• I have disclosed AI assistance in the PR description.
• I confirm that I have personally reviewed and understood all AI-generated content, and accept full responsibility for it.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!