Implementation of BlobLoader with S3

[eyurtsev]

An implementation of BlobLoader with S3 would be fantastic. It'll allow users to easily hook their s3 data to langchain parsers and indexing code.

Here's an explanation about the abstractions:
https://python.langchain.com/docs/modules/data_connection/document_loaders/custom/#blob-loaders

The minimal feature set for this to be useful:

1. filters by keys matching a given prefix
2. filter for last_modified_date
3. filter for max file size
4. filter for file extension
5. filter for min file size

Additional features:
7. if available filter for file mimetype