vllm-omni support

**Is your feature request related to a problem? Please describe.**
I am trying to get tritonserver working with vllm-omni based model(Qwen3_tts)
**Describe the solution you'd like**
The vllm backend to support omni