Add part_of for discovered families of prefixes

[bgyori]

This PR adds part_of relationships between prefixes and their assumed parents where missing. The PR stems from the discussion in #1222. I did the following:

• Group prefixes by their root (e.g., aop for aop.events, aop.stressor, etc.) into families. Spot checking these relations show that these appear to be genuine families of prefixes, though there could be exceptions.
• Check which elements of a family already have a part_of relationship to the root
• Under the condition that a member doesn't have a part_of relationship to a root and the root is also a Bioregistry prefix, add a part_of relationship from the given prefix to the root prefix.

Questions to consider:

• We could extend this to add part_of relations to parents that aren't Bioregistry entries. An example that already exists is pubchem.
• Is the code that made these changes useful at all to be in version control? It might just be a one-off so I am not immediately adding it anywhere.
• Are there any potential concerns where a part_of added here isn't actually correct, i.e., where two prefixes with .-s in them share a root but shouldn't actually have a part_of relationship to the same root?
• As https://biopragmatics.github.io/bioregistry/datamodel/#part-of points out, there are family relationships other than the ones following the . separation pattern which aren't picked up here - these might require manual curation.

[bgyori]

The test failure appears to have to due with a conflict between has_canonical and part_of relationships. In particular, insdc.run is annotated as has_canonical for ena.embl but when adding insdc.run-part_of- insdc, it results in insdc.run not being surfaced in the top-level of the registry anymore which is what the test appears to expect.