feat(audio): batch Qwen3-ASR transcriptions

[qinxuye]

Summary

• batch concurrent Qwen3-ASR transcription requests through the model's native list inference
• preserve per-request language and response formatting while grouping compatible inference kwargs
• harden the shared batch worker against cancellation, batch-size overflow, and result-count mismatches

Why

Qwen3-ASR supports batched transcription, but Xinference currently invokes each concurrent audio request separately. This change lets compatible requests share one model call while preserving the existing API response behavior.

Validation

• pytest -q xinference/model/audio/tests/test_qwen3_asr.py
• pre-commit run --files xinference/model/audio/qwen3_asr.py xinference/model/audio/tests/test_qwen3_asr.py xinference/model/batch.py

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!