Implement sidecar import functionality

[dglazkov]

An idea for importing book-like material: long documents that have a "sidecar" of references that aren't contained in the document

• first, we chunk a document into a Google spreadsheet
• then, someone can go over the spreadsheet and add cells for title, url, image_url, description for each chunk
• and then we import the spreadsheet

[dglazkov]

Maybe like this:

• [ ] Implement convert.out csv <library_file> that spits a library as a CSV file (with or without embeddings?)
• [ ] Implement Google Spreadsheet importer