Fix mixed timestamp timezone parsing

[Gambit-Checkmate]

What & why

A transcript mixing offset-aware and offset-naive timestamps crashes when its runs are sorted. Parse timestamps without offsets as UTC, while preserving explicit offsets, so sorting and duration subtraction use comparable datetimes independently of the local timezone. Make the missing-timestamp sort fallback UTC-aware as well.

Closes #27. Also addresses the missing-timestamp sorting crash described in #3.

Changes

• Document the UTC assumption for timestamps without an offset.
• Add transcript-level regressions for mixed naive/aware timestamps, explicit nonzero offsets, missing timestamps, and malformed timestamps. Three cases failed with TypeError before the fix; the explicit-offset case already passed.
• Verify chronological ordering and a 30-second duration across different timestamp representations.

Validation

All 33 tests pass. Both modified Python files pass the format check; the new test file passes Ruff. Ruff on parse.py retains two pre-existing diagnostics (UP035 and FURB162), verified against origin/main; the previous DTZ901 diagnostic is resolved. git diff --check passes.

Checklist

• Focused change with a clear description
• Ran format / lint / tests (existing lint diagnostics noted above)
• Updated docs
• Linked related issues

[github-actions]

Merged. Thanks a lot, @Gambit-Checkmate.

Agentrace gets better every time someone outside the repo digs in like this.

If you find Agentrace useful, consider dropping a star. It helps more people find it.

Happy to see more from you whenever you have the time.