Fix broken file transcription and speaker diarization

qwen3 backend: model list pointed at nonexistent HF repos (Qwen3-ASR-2B/8B
don't exist), and inference went through the generic HF ASR pipeline with
Whisper-only options (return_timestamps, task/language generate_kwargs) that
Qwen3-ASR's chat-style architecture doesn't support. Switched to the real
0.6B/1.7B-hf checkpoints and drive them via
processor.apply_transcription_request() + model.generate(). Also added an
ffmpeg pre-conversion step since the model's feature extractor can't decode
m4a via librosa.

faster-whisper backend: speaker diarization was broken by two pyannote.audio
API changes it hadn't caught up with (use_auth_token= renamed to token=, and
pipeline() now returns a DiarizeOutput wrapper instead of an Annotation
directly). Also pinned LD_LIBRARY_PATH for that worker so it picks up its own
venv's cuDNN instead of the older one shadowing it via the container's global
LD_LIBRARY_PATH, which crashed pyannote's GPU init.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

app/asr/workers/faster_whisper_worker.py

@@ -209,7 +209,7 @@ def _get_diarization_pipeline():
             raise RuntimeError("화자 분리를 사용하려면 PYANNOTE_HF_TOKEN 환경변수를 설정하세요.")
         _DIARIZATION_PIPELINE = Pipeline.from_pretrained(
             "pyannote/speaker-diarization-3.1",
-            use_auth_token=PYANNOTE_HF_TOKEN,
+            token=PYANNOTE_HF_TOKEN,
         )
         if DEVICE == "cuda":
             import torch
@@ -234,7 +234,9 @@ def _apply_diarization(
         if max_speakers is not None:
             kwargs["max_speakers"] = max_speakers
 
-    diarization = pipeline(str(audio_path), **kwargs)
+    # pyannote.audio 4.x: pipeline() returns a DiarizeOutput dataclass instead of
+    # an Annotation directly; the itertracks-capable Annotation is .speaker_diarization.
+    diarization = pipeline(str(audio_path), **kwargs).speaker_diarization
     segments = result.get("segments", [])
     for seg in segments:
         seg_start = float(seg.get("start", 0))