fixing error stt and tts, empty chunk audio

Fix: load Google credentials from env var for HuggingFace Spaces deployment

Add creds, Update used TTS model

Feat: update stt chirp3

Feat: STT & TTS Gemini, Creds Gemini Cloud Speech, Generate Buffer Audio, Testing

TTS and STT using gemini model

[NOTTICKET] add gemini client

[NOTICKET] feat: endpoint stt and tts

[KM-501, KM-498, KM-499] Web Socket, STS and TTS