feat: add newer model tts chatterbox to ai-runner text-to-speech

[JJassonn69]

This PR adds the feature for text-to-speech using newer ResembleAI/Chatterbox which as voice cloning feature.

So as not to cause change in the go-livepeer it uses the same application/json format to relay the information between orchestrator and runner.

You can build the text-to-speech docker image from the Dockerfile.text-to-speech similar to any other images.

Docker build -t livepeer/ai-runner:text-to-speech -f Dockerfile.text-to.speech .

for testing purpose you can access the uvicorn server after starting the container with the pipeline and model

docker run --name text-to-speech -e PIPELINE=text-to-speech -e MODEL_ID=chatterbox --gpus all -p 8000:8000 -v ./models:/models <docker-image-build-above>

In the uvicorn server, you will see params for model_id, text and prompt_audio_base64

model_id: ResembleAI/Chatterbox
text:
prompt_audio_base64:

The text converted to base64 looks like UklGRjiTIQBXQVZ.....AADUAADQAADAAACkAAA==

Here is a attached file with the base64 converted audio from my voice.
audio_prompt_base64.txt