Add subsetting functionality

[cthoyt]

This functionality is useful for downstream applications like the following:

1. You load a comprehensive extended prefix map, e.g., from the Bioregistry using curies.get_bioregistry_converter().
2. You load some data that conforms to this prefix map by convention. This is often the case for semantic mappings stored in the SSSOM format
3. You extract the list of prefixes actually used within your data
4. You subset the detailed extended prefix map to only include prefixes relevant for your data
5. You make some kind of output of the subsetted extended prefix map to go with your data. Effectively, this is a way of reconciling data. This is especially effective when using the Bioregistry or other comprehensive extended prefix maps.

Here's a concrete example of doing this (which also includes a bit of data science)
to do this on the SSSOM mappings from the Disease Ontology project.

>>> import curies
>>> import pandas as pd
>>> import itertools as itt
>>> commit = "faca4fc335f9a61902b9c47a1facd52a0d3d2f8b"
>>> url = f"https://github.com/mapping-commons/disease-mappings/blob/{commit}/mappings/doid.sssom.tsv"
>>> df = pd.read_csv(url, sep="\t", comment="#")
>>> prefixes = {
...     curies.Reference.from_curie(curie).prefix
...     for column in ["subject_id", "predicate_id", "object_id"]
...     for curie in df[column]
... }
>>> converter = curies.get_bioregistry_converter()
>>> slim_converter = converter.get_subconverter(prefixes)

This PR also sneaks in a related documentation update to pandas dataframe processing