Refactor TTS worker text prep for v1.9.2

[amariichi]

Summary

• move spoken-text cleanup out of face-app and into worker shared prep
• keep backend-specific shaping in the selected TTS engine
• move the audio replay button left to avoid overlapping the Esc control
• bump repo package versions to 1.9.2 for the release

Validation

• npm test
• uv run --project tts-worker python -m unittest discover -s tts-worker/tests