Add ability to retrieve and cache prefix maps

Currently the code makes some assumptions about prefix expansions

We would like it such that callers at any level (python, cli) have full control, and can specify either

- custom prefix-URL pairs, e.g as TSVs or YAML
- named prefix maps

named prefix maps are just custom names for bags of prefix expansion tuples.

There would be a few hardcoded named prefix maps:

- obo
- semweb

but users should be able to name their own.

bioregistry can be used to allow users to download standard prefixes, but named prefix maps would be cached using pystow to eliminate runtime network dependencies. The hardcoded prefix maps like OBO could potentially be bundled such that behavior is deterministic and coupled to the oaklib version

Start by tracing the existing prefix-expansion assumptions in the Python library and CLI entry points. Review how bioregistry and pystow could fit the requested flow, then define completion as support for custom TSV/YAML pairs, user-named maps, cached maps, and deterministic OBO and semweb maps without runtime network access.

Written by the indexing model from the issue text.