fix(trickle): migrate wait_for → asyncio.timeout to fix cancellation masking

[rickstaa]

Modernization, not a bug fix

This PR migrates four asyncio.wait_for callsites to asyncio.timeout (Python 3.11+). The migration is correct and worthwhile on its own — asyncio.timeout was added specifically to address wait_for's cancellation behavior — but it does not fix the shutdown-noise symptom originally tracked in #10.

After migration, live runner E2E (examples/runner/live_grayscale/test.sh) still reproduces the same TrickleSegmentWriteError and "close suppressed" log lines. Verified via timestamped logs that the timeouts firing are legitimate — the segment's HTTP post-body consumer is dead/wedged, so queue.put blocks until the 5s deadline. That's a separate root cause; #10 has been updated to describe it.

Why migrate anyway

asyncio.wait_for has documented races where outer cancellation can be masked as TimeoutError (bpo-32751). We didn't actually hit that race in our reproduction, but asyncio.timeout uses task.uncancel() internally to handle cancellation correctly by design. Pre-emptive correctness — and it's the recommended primitive for new code from 3.11 onward.

Changes

File	Change
trickle_publisher.py	SegmentWriter.write/close migrated. Explicit CancelledError re-raise in close() so BaseException doesn't log a misleading "close suppressed" warning when cancel propagates.
media_publish.py	Idle-timeout watchdog (line 945) and drain-on-shutdown (line 1049) migrated.
pyproject.toml	requires-python bumped from >=3.10 to >=3.11. Required because asyncio.timeout is 3.11+. Docker and CI already run 3.11; 3.10 EOL is October 2026.

Validation

• All four asyncio.wait_for callsites in src/livepeer_gateway/ migrated; only generated gRPC wait_for_ready parameters remain (unrelated).
• Imports clean.
• examples/runner/live_grayscale/test.sh still passes (and still produces the unrelated shutdown traceback noise tracked in TricklePublisher.SegmentWriter masks CancelledError as TrickleSegmentWriteError on shutdown #10).

Out of scope

Fixing the shutdown noise — see updated #10 for the actual root cause and proposed approach.