Add Mid-Generation Streaming Support

[cvaz1306]

Summary

This PR introduces a high-performance streaming architecture to the TTS server. Users can now hear and see audio as it is being generated chunk-by-chunk, significantly reducing "Time to First Sound" (TTFS) for long texts. This PR also includes critical stability fixes for audio degradation and digital noise.

Key Features

• Added a new /tts/stream API endpoint that yields audio chunks immediately using an asynchronous generator.
• A new UI control allows users to switch between standard batch generation and live streaming mode.

Testing Instructions

1. Restart the server.
2. Enter a long paragraph of text (3-4 sentences).
3. Ensure "Stream Audio" is checked.
4. Click "Generate Speech".
5. Verify: Audio should start playing within seconds.
6. Verify: Press the "Pause" button during generation to ensure audio stops and can be resumed.