Epic: reduce dependencies on central sqlite database

Reason:

- sync: the semsql builds always lag the ontology release
- reduce egress charges

Currently it's highly attracting to use the sqlite adapter because

- it is fast
- it is highly feature complete

The first task should be to make some of the other adapters more feature complete so it is more attractive to use them. This might include those that use an external service (eg OLS, Ubergrpaph); or now that we have better OWL support though py-hornedowl that should become the default for more things.

cc @kltm

* https://github.com/INCATools/semantic-sql/issues/110
* https://github.com/linkml/linkml-term-validator/issues/51
* #894

Start by comparing the SQLite adapter with the other adapters mentioned in the issue, including OLS, Ubergraph, and py-hornedowl, and review the linked issues for existing feature gaps. Done means the selected non-SQLite adapters provide enough of the needed functionality that the central SQLite dependency is reduced.

Written by the indexing model from the issue text.