[azopenai] Whisper support

[richardpark-msft]

Just running into some API generation complications, want to track the work.

The basics are that we end up with two new functions - GetAudioTranscription and GetAudioTranslation, which are very similar. Both of them accept a set of inputs like the prompt, the model and the actual audio data to be translated/transcribed.

So far, the default generation is working "ok" but it looks like some fidelity is being lost since we actually do have multipart support in our autorest.go. So:

• The requests should be made to be multipart/form-data instead of defaulting to application/json
• The return values are either plain text or a structured JSON body. The generated code is not recognizing that.

Still haven't gotten it all the way through, but these are some of the immediate things.

[richardpark-msft]

Fixed with #21599