Add data integrity tests for unique IRIs

[cthoyt]

This PR adds tests to ensure that when we generate a prefix map using the Bioregistry that all of the IRIs are unique. There are two scenarios we can use to mitigate this I can think of offhand:

1. When entries are in parent/child relationship, like kegg with kegg.compound, kegg.drug, etc. we can enforce that the part_of annotation is used for all of the children. If only one prefix in a group of prefixes with the same IRI doesn't have a part_of relationship, we'll assume it's the parent and default to using it.
2. When entries are annotated with a provides, then they will automatically not be considered in the prefix map, such as ctd.gene, which provides for ncbigene

When neither of these work, this PR has also introduced the has_canonical relationship between prefixes that maps one prefix to another.

There are some entries that seem to be duplicates of each other like glycomedb and glytoucan. In this case, glytoucan is the actual name of the database so glycomedb gets the "has_canonical": "glytoucan", denoting it is higher priority (see identifiers-org/identifiers-org.github.io#167). I wasn't confident enough to merge them, but the wb and wormbase entries are now indeed merged.

The tests enabled exhaustive curation of scenarios 1, 2, and the pmapto is used for the rest.

[cthoyt]

@matentzn FYI this is the big step necessary in improving the curation in the Bioregistry to support generating prefix maps that are guaranteed to be bidirectionally unique :)