Gemini: support TTS models and configuration

### Privileged issue

- [x] I am a LangChain maintainer, or was asked directly by a LangChain maintainer to create an issue here.

### Issue Content

See https://ai.google.dev/gemini-api/docs/speech-generation
* Support response modalities setting
* Add SpeechConfig / VoiceConfig
* Verify audio output works
* Test with both gemini-2.5-flash-preview-tts and gemini-2.5-pro-preview-tts