Fix spoken audio responses on Gaudi and add tests

[dmsuehir]

Description

During MMQnA phase 3 testing on Gaudi, I found that audio responses are not working on Gaudi:

$ curl http://${host_ip}:${MEGA_SERVICE_PORT}/v1/multimodalqna \
        -H "Content-Type: application/json" \
        -d '{"messages": [{"role": "user", "content": [{"type": "text", "text": "hello, "}, {"type": "image_url", "image_url": {"url": "https://www.ilankelman.org/stopsigns/australia.jpg"}}]}, {"role": "assistant", "content": "opea project! "}, {"role": "user", "content": "chao, "}], "max_tokens": 10, "modalities": ["text", "audio"]}'
Internal Server Error

I debugged this and found that the TTS endpoint was incorrect due to a missing env var. Also, we don't have any megaservice tests with "modalities": ["text", "audio"], so I added one for both Xeon and Gaudi.

Issues

Fix for a MMQnA phase 3 feature

Type of change

List the type of change like below. Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)

Dependencies

N/A

Tests

Added a test for getting an audio response

[mhbuehler]

LGTM