Fix stale state in user turn stop strategies between turns

[markbackman]

Context

The issue filed in #4053 highlights a case where state can bleed over between one turn to the next. This PR attempts to clean up the state between turns to ensure we don't have any state carrying over.

Summary

• Fixed premature user turn stops caused by late transcriptions arriving between turns. When a TranscriptionFrame arrived after a turn ended but before the next one started, its text would persist as stale state into the next turn, triggering an immediate stop before the current turn's real transcript arrived.
• Stop strategies are now reset at both turn start and turn stop, ensuring no state leaks across turn boundaries. The if self._user_turn: return guard in _trigger_user_turn_start ensures this only fires once per turn, not on mid-turn VAD restarts (speak-pause-speak).
• Pending timeout tasks in SpeechTimeoutUserTurnStopStrategy and TurnAnalyzerUserTurnStopStrategy are now cancelled during reset() to prevent stale timeouts from firing after a turn boundary.

Testing

uv run pytest tests/test_user_turn_stop_strategy.py tests/test_user_turn_controller.py -v

Fixes

• Fixes Stale _text in TurnAnalyzerUserTurnStopStrategy causes 'one turn behind' bug #4053

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
...ser_stop/speech_timeout_user_turn_stop_strategy.py	91.11% <100.00%> (+0.30%)	⬆️
...user_stop/turn_analyzer_user_turn_stop_strategy.py	79.27% <100.00%> (-0.36%)	⬇️
src/pipecat/turns/user_turn_controller.py	95.16% <100.00%> (+0.07%)	⬆️

... 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]

Nice catch! LGTM!