set audio_column to audio instead of path

The task template in DatasetInfo specifies the audio_column as "path". This leads to an error when trying to load the dataset. Changed the audio_column to "audio" so that dataset can be loaded without datasets library complaining.

nst.py

@@ -91,7 +91,7 @@ class NSTDataset(datasets.GeneratorBasedBuilder):
             supervised_keys=None,
             homepage=_URL,
             task_templates=[
-                AutomaticSpeechRecognition(audio_column="path", transcription_column="text")
+                AutomaticSpeechRecognition(audio_column="audio", transcription_column="text")
             ],
         )