Update dataset structure to 3 shards and add final parquet file

Successfully merged and synchronized audio features schema

Fix dataset viewer error by auto-casting via datasets library

Upload data/train-00000-of-00001.parquet with huggingface_hub