Keeping audio and clock queue in sync.

[filipi87]

Summary

• When a TTSStoppedFrame carries a presentation timestamp (pts), it is ow enqueued in both the audio_queue and the clock_queue so that timed frames (e.g. word-boundary events) and audio playback stay in sync.
• When the audio task finishes sending all preceding audio and dequeues the TTSStoppedFrame, it signals the clock task via _clock_flush_event to stop waiting on timestamps and drain all pending frames immediately.
• The clock task signals back via _clock_drained_event once the drain is complete, so the audio task can push TTSStoppedFrame downstream.
• A BOT_VAD_STOP_FALLBACK_SECS timeout guards the wait on _clock_drained_event to prevent a deadlock if the clock task fails before completing the drain.

[codecov]

Codecov Report

❌ Patch coverage is 0% with 26 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/pipecat/transports/base_output.py	0.00%	26 Missing ⚠️

Files with missing lines	Coverage Δ
src/pipecat/transports/base_output.py	17.67% <0.00%> (-1.05%)	⬇️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[filipi87]

I tested all TTS services that provide word timestamps:

• Rime
• ElevenLabs
• Inworld
• Gradium
• Hume
• ResembleAI
• Azure
• Cartesia

Only Rime is returning incorrect timestamps.

However, the fix in the PR is not specifically addressing the Rime issue. Instead, it acts as a safeguard to ensure both queues stay in sync by using the TTSStopped frame as a gatekeeper.

But closing it for now, since Rime is currently the only service that would benefit from this change.