Audio asset with pydub conversion fails to write on S3

[lhoestq]

See https://huggingface.co/datasets/zeio/auto-pale/discussions/1 (ogg audio files)

File \"/src/libs/libcommon/src/libcommon/viewer_utils/asset.py\", line 111, in create_audio_file
   segment.export(f, format=suffix[1:])
File \"/src/services/worker/.venv/lib/python3.9/site-packages/pydub/audio_segment.py\", line 868, in export
   out_f.seek(0)
File \"/src/services/worker/.venv/lib/python3.9/site-packages/fsspec/spec.py\", line 1737, in seek
   aise OSError(ESPIPE, \"Seek only available in read mode\")
OSError: [Errno 29] Seek only available in read mode

we might have to write to a BytesIO buffer before writing the audio actual audio asset file, since fsspec doesn't allow seek for wb files somehow.

[AndreaFrancis]

Closed by #2085
Also https://datasets-server.huggingface.co/rows?dataset=zeio%2Fauto-pale&config=pulled&split=train&offset=0&length=100 looks to be working now.