Audio: Running an audio model sometimes runs the fastchat_server without it being selected

[deep1401]

I tried to run the Orpheus model and this is what it did on its own:

## Plugin unsloth_text_to_speech_server installed successfully.
🏃 Loading Inference Server for canopylabs/orpheus-3b-0.1-ft with {'inferenceEngine': 'fastchat_server', 'inferenceEngineFriendlyName': 'Fastchat Server'}
Starting Standard FastChat Worker