Fix Time Travel stale index.lock cleanup and await progress updates

[zayzay109]

Summary

This PR fixes two small stability issues observed in Agent Zero runtime paths:

1. Time Travel can leave a stale repo.git/index.lock when a Git subprocess is killed by subprocess.TimeoutExpired, poisoning later snapshots for that workspace.
2. code_execution_tool calls async Tool.set_progress() without await in two output-progress paths, producing coroutine warnings and potential progress/update instability under load.

Changes

Time Travel

Adds a best-effort _cleanup_git_index_lock() helper and calls it when Git subprocesses time out in:

• _run_git_dir()
• _git()
• _git_bytes()

The timeout exception is re-raised. The patch does not change GIT_TIMEOUT_SECONDS, does not suppress errors, and does not alter normal successful Git behavior.

Code execution tool

Changes two bare calls:

self.set_progress(truncated_output)

to:

await self.set_progress(truncated_output)

Why

If a Time Travel Git command times out, the killed Git process may leave index.lock behind. Later snapshot attempts then fail with:

fatal: Unable to create '.../repo.git/index.lock': File exists.

Cleaning the lock after timeout prevents one failed snapshot from permanently poisoning that workspace.

The set_progress calls are in async functions and should await the coroutine.

Validation

Validated locally with:

python -m py_compile plugins/_time_travel/helpers/time_travel.py
python -m py_compile plugins/_code_execution/tools/code_execution_tool.py

Also validated in a live Agent Zero environment with a Time Travel stability guard after remediation:

{
  "verdict": "PASS",
  "workspace_count": 9
}

No current stale Time Travel index.lock files remained after the fix.

Notes

This is intentionally minimal defensive cleanup. It does not disable Time Travel, change timeout values, or add environment-specific behavior.