Provide the ability to extract all mappings between multiple sources

[gaurav]

This has been requested by a bunch of people, and with the new Corcord.parquet file it shouldn't be too hard to implement here.

The idea is to run something like:

$ uv run babel-xrefs dump --prefix CHEBI --prefix PUBCHEM.COMPOUND --prefix CAS --format tsv

And to produce all of the cross-references between those prefixes in a format that can be shared.

This has been requested at:

• Add a bulk mapping download API endpoint NCATSTranslator/Babel#641
• Export big subsets out of Babel by prefix NCATSTranslator/Babel#640