Switch to Whelk reasoner

[gouttegd]

This PR makes CL use the Whelk reasoner for all its build and test pipelines.

closes #2717

[gouttegd]

@dosumis

Here’s a diff on “CL-base with ELK” (“left” ontology) versus “CL-base with WHELK” (“right” ontology):

2 axioms in left ontology but not in right ontology:
- SubClassOf(<http://purl.obolibrary.org/obo/CL_0000557> <http://purl.obolibrary.org/obo/CL_0011026>)
- SubClassOf(<http://purl.obolibrary.org/obo/CL_0001029> <http://purl.obolibrary.org/obo/CL_0000763>)

4 axioms in right ontology but not in left ontology:
+ SubClassOf(<http://purl.obolibrary.org/obo/CL_0000995> <http://purl.obolibrary.org/obo/CL_0011026>)
+ SubClassOf(<http://purl.obolibrary.org/obo/CL_0001012> <http://purl.obolibrary.org/obo/CL_0011026>)
+ SubClassOf(<http://purl.obolibrary.org/obo/CL_0001019> <http://purl.obolibrary.org/obo/CL_0000763>)
+ SubClassOf(<http://purl.obolibrary.org/obo/CL_0001030> <http://purl.obolibrary.org/obo/CL_0011026>)

[dosumis]

Hypothesis - WHELK is able to reason with disjunction and so can take advantage of equivalentClass Defs with OR. ELK is not. This is consistent with doc

• every case above appears to involve reasoning from disjunction in the superclass position, either directly or indirectly (the two removed axioms appear to be removed because they are redundant with inference on a superclass defined using OR)

Cautionary note - I'm seeing this error when running WHELK on CL in Protege

CC @balhoff - is this error a potential problem?

[balhoff]

@gouttegd that error is expected. I might ought to handle that a different way but it doesn't seem to bother Protege.