fix(agent): guard finalize_turn cleanup chain so it never drops the response

[Megashubham]

What does this PR do?

When a turn hits max_iterations, finalize_turn ran three unguarded cleanup steps after the model's summary:

• _save_trajectory (file I/O)
• _cleanup_task_resources (remote VM/browser teardown)
• _persist_session (SQLite write)

Any exception from these propagated out of run_conversation, silently discarding final_response. Subprocess wrappers saw empty stdout with no traceback.

Each step is now wrapped in its own try/except so one failure cannot skip the others. Failures log at ERROR with a traceback and are surfaced on the result dict via cleanup_errors; the partial response is always returned. Clean turns never get a cleanup_errors key.

Related Issue

Fixes #8049

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✅ Tests (adding or improving test coverage)

Changes Made

• agent/turn_finalizer.py: wrap each of the three core cleanup steps in its own try/except, collect errors in _cleanup_errors list, surface via cleanup_errors on result dict
• tests/agent/test_turn_finalizer_cleanup_guard.py: regression coverage — all-three-raise, each-one-raises-in-isolation (others still run), and clean-turn-no-key

How to Test

1. Run pytest tests/agent/test_turn_finalizer_cleanup_guard.py -v — 5 tests pass
2. E2E: create a stub agent where _save_trajectory raises; before this fix run_conversation returns no response; after this fix the response is preserved and cleanup_errors shows the error.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix
• I've run pytest tests/agent/test_turn_finalizer_cleanup_guard.py -v and all 5 tests pass
• I've added tests for my changes
• I've tested on: macOS

Documentation & Housekeeping

• N/A — no config keys or tool schemas changed

[alt-glitch]

Duplicate of #50009 — same fix for #8049, filed ~1.5h earlier. Both wrap the three post-loop cleanup steps (_save_trajectory, _cleanup_task_resources, _persist_session) in independent try/except blocks, collect a _cleanup_errors list, surface it via a cleanup_errors result key, and add the identically-named regression test tests/agent/test_turn_finalizer_cleanup_guard.py. Related: #43057 and #46183 (same code path, different approach).