qwen3-asr.cpp OpenAI-compatible server

POST /v1/audio/transcriptions
  multipart/form-data fields:
    file=@audio.wav
    model=qwen3-asr
    language=<optional>
    response_format=json|text|verbose_json