Add (Azure) speech capabilities (STT / TTS)

[iMicknl]

Would be great to add (optional) voice capabilities to the sample app, based on a feature flag. Similar to what is currently available in the Azure OpenAI Playground.

Components

• Azure Speech to Text (after click / hold on input button)
• Azure Text to Speech (on streaming output, not just on the final answer)

Requires:

• Azure Speech deployment
• Back-end work to retrieve token (Managed Identity or Key) and pass a short-term access token
• Environment variables to enable / disable voice, set voice configuration

Open discussion

• How to set the input language, via environment variables or in the interface?
• Should we use recognizeOnceAsync (one-shot) or continuous until user closes / a long pause is detected. Benefit of continous is that you can stream the input to the question field, where with one-shot you can only share the result.

[pamelafox]

If its helpful, the speech team contributed text-to-speech in a PR for our similar RAG repository here: https://github.com/Azure-Samples/azure-search-openai-demo/ The comments on that PR might also be helpful to see what developers have requested from the feature.