perf: cap torch threads per camera worker

Default TORCH_NUM_THREADS=1 so 14-20 workers do not each spawn a
28-thread BLAS pool.

run_worker.py

@@ -38,6 +38,7 @@ def main() -> int:
     ap.add_argument("--enhance-chunk-s", type=float, default=1.0)
     ap.add_argument("--enhance-hop-s", type=float, default=0.2)
     ap.add_argument("--beamform-mode", default="auto")
+    ap.add_argument("--torch-num-threads", type=int, default=1)
     args = ap.parse_args()
 
     cam = json.loads(args.camera_json)
@@ -47,8 +48,9 @@ def main() -> int:
     )
 
     from config import Config  # noqa: F401  (ensures .env is loaded)
+    from audio_cleanup import AudioCleaner, apply_torch_thread_limits
+    apply_torch_thread_limits(args.torch_num_threads)
     import livekit.rtc as rtc
-    from audio_cleanup import AudioCleaner
 
     identity = cam["identity"]
     video_dev = cam["video_device"]
@@ -62,6 +64,7 @@ def main() -> int:
             chunk_s=args.enhance_chunk_s,
             hop_s=args.enhance_hop_s,
             beamform=args.beamform_mode,
+            torch_num_threads=args.torch_num_threads,
         )
         log.info("audio backend: %s", cleaner.backend)