clean up on task cancelation to avoid resource leakage

[rasdani]

Description

related to fix #793
avoids resource leakage when canceling rollout tasks.

in the case of task cancellation this could call self._cleanup() twice and thus assumes some idempotency.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Test improvement

Testing

• All existing tests pass when running uv run pytest locally.
• New tests have been added to cover the changes

Checklist

• My code follows the style guidelines of this project as outlined in AGENTS.md
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Any dependent changes have been merged and published

Additional Notes

---

Note

Ensures resources are cleaned up when rollouts are cancelled and clarifies cleanup expectations in docs.

• In MultiTurnEnv.rollout, catch asyncio.CancelledError and call self._cleanup(state) before re-raising; minor restructuring of the setup/loop error handling
• Docs (environments.md, AGENTS.md, reference.md) now stress that @vf.cleanup/@vf.teardown methods must be idempotent and handle errors gracefully

^{Written by Cursor Bugbot for commit 9f5c106. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}