feat(elevenlabs): add eleven_v3 model support with HTTP streaming fallback

[yusuf-eren]

Adds support for ElevenLabs eleven_v3 TTS model by implementing HTTP streaming fallback. The plugin automatically detects when eleven_v3 is used and routes to the HTTP /stream endpoint instead of WebSocket.

Changes

• Add HTTPSynthesizeStream class for HTTP-based streaming
• Automatically detect and route eleven_v3 to HTTP streaming
• Add comprehensive test coverage (8 unit tests)
• Include example demonstrating usage

Implementation

tts = elevenlabs.TTS(model="eleven_v3")
stream = tts.stream()  # Automatically uses HTTP streaming

Other models continue using WebSocket unchanged.

Limitations

HTTP streaming has higher latency (~1.5s TTFB) compared to WebSocket models (~300ms) due to text buffering requirements. This is an ElevenLabs API limitation since eleven_v3 doesn't support WebSocket connections.

Fixes 403 errors when using eleven_v3.