Add parents for implicitly declared classes

[cthoyt]

Motivated by biopragmatics/bioregistry#1811, follow-up to #486.

This PR adds a function that can be used to search all terms in an ontology, identify external references that weren't explicitly declared as a class in the ontology, and add a declaration.

All terms declared this way get put under an ontology-wide ad hoc grouping term, which in turn has children corresponding to each prefix for which undeclared classes were found.

For example, in HGNC, the ontology-wide grouping term is http://purl.obolibrary.org/obo/hgnc#aux, under which are terms like http://purl.obolibrary.org/obo/hgnc#aux-ec (fo

Here's how it looks for HGNC:

There are a few caveats to this approach:

1. Axiom injection (🤷)
2. Doesn't get nice metadata like labels
3. Weird looking labels because there are no labels (will be improved with Improve guessing of names protegeproject/protege#1322)

[matentzn]

All terms declared this way get put under an ontology-wide ad hoc grouping term, which in turn has children corresponding to each prefix for which undeclared classes were found.

From a UI perspective this sounds sane for now, but the problem is that this will fold these metaclasses (they are not real domain classes!) into our ontologies downstream. So when we extract a module we will get these injected classes through the pyobo ingests, and they will end up in our imports unless we explicitly exclude them. I think this is probably not wise, but probably not a desaster especially if these could be materialised by default only in the non-default release product..