Validation of HAS ONLY queries

Quite a few databases now implement the optional "HAS ONLY" filter construct, though it seems that many do it incorrectly (like we did until @JPBergsma fixed it in #810 and #861). We should try to add a test for it in the validator (if the server implements it) - unfortunately it will be a bit tricky to construct test queries as the current failure mode is that "x HAS ONLY A, B, C" does not return results where x is a subset of {A, B, C}, so we will an additional query to that any entries exist with only {A, B} (say) and check that they are also returned in the final HAS ONLY query.

Typed this in a hurry, but hope that makes sense! I am also writing this issue so that I remember to bring it up at the next OPTIMADE meeting.

Start at the validator and review the HAS ONLY fixes referenced in #810 and #861, then examine how existing validation tests construct queries. Add coverage that checks subset results, including an additional query confirming entries with only {A, B}, and verify the test works when the server implements HAS ONLY.