fix(transcript_mirror): swallow CancelledError in eager-flush done callback

[SuperMarioYL]

Summary

Fixes #930.

The eager-flush task in TranscriptMirrorBatcher.enqueue is registered with

self._flush_task.add_done_callback(lambda t: t.exception())

The intent is to retrieve the task's exception so asyncio doesn't warn about an unretrieved one, but on a cancelled task Task.exception() itself raises CancelledError (Python 3.8+). asyncio's callback machinery then logs

ERROR asyncio: Exception in callback ... <lambda>(<Task cancell...>)
  File ".../transcript_mirror_batcher.py", line 91, in <lambda>
    self._flush_task.add_done_callback(lambda t: t.exception())
                                                 ^^^^^^^^^^^^^
asyncio.exceptions.CancelledError

every time an in-flight eager-flush task is cancelled — i.e. on every event-loop teardown with pending eager flushes, and whenever Query.close() cancels child tasks while a drain is mid-flight. Functionality is unaffected (the inner _drain handles cancellation correctly via async with self._lock); the bug is pure log noise but it pollutes test output and production logs for users running with session_store_flush=\"eager\".

Fix

Replace the lambda with a small named _swallow_done_exception helper that skips cancelled tasks before calling .exception():

def _swallow_done_exception(task: asyncio.Task[None]) -> None:
    if task.cancelled():
        return
    task.exception()

Equivalent to the inline form lambda t: None if t.cancelled() else t.exception() but names the intent.

Test

Added TestTranscriptMirrorBatcher.test_eager_flush_cancellation_does_not_log_callback_error:

• Builds a batcher with max_pending_entries=0 so any enqueue schedules an eager flush.
• Uses a hanging store so the eager flush blocks indefinitely.
• Cancels the in-flight flush task and yields the loop so the done callback fires inside the caplog.at_level("ERROR", logger="asyncio") block.
• Asserts the asyncio logger recorded no Exception in callback records at ERROR level.

Verified the test fails on master (one CancelledError record) and passes with the fix.

Validation

• python -m ruff check src/ tests/ — clean
• python -m ruff format --check src/ tests/ — clean
• python -m mypy src/ — clean (Success: no issues found in 24 source files)
• python -m pytest tests/ — 746 passed, 5 skipped (full suite)

Notes

• Pure bug fix; no behavior change beyond log silence on cancellation.
• The outer try/except in _drain and close is untouched.
• Issue Flaky tests: eager-flush transcript-mirror tests fail with 2 sleep(0) yields (test_transcript_mirror.py) #928's Captured log teardown traceback (referenced in Bug: TranscriptMirrorBatcher.enqueue's add_done_callback lambda raises CancelledError when task is cancelled #930) is the same bug's symptom; this fix should clear that log noise from those tests as well.