Add torchcodec dependency for torchaudio audio loading

Newer torchaudio (2.6+) defaults to the torchcodec backend for
audio loading. Without it, torchaudio.load() fails with ImportError
when processing .opus files during auto-labeling.

requirements.txt

@@ -16,6 +16,7 @@ omegaconf>=2.3.0
 hydra-core>=1.3.0
 
 # Audio Processing
+torchcodec>=0.1.0
 librosa>=0.10.0
 soundfile>=0.12.0
 scipy>=1.11.0