Support loading/streaming data from s3 locations

[nihaljn]

This is what's supported at the moment:

• Local - JSON, JSONL, HuggingFace parquet
• Remote
  • HuggingFace Hub - HuggingFace parquet

Supporting S3 loading/streaming would entail supporting the following:

• Local - JSON, JSONL, HuggingFace parquet
• Remote
  • HuggingFace Hub - HuggingFace parquet
  • S3 - JSON, JSONL, HuggingFace parquet