Tests currently include lexical values in EntityRef slots

[matentzn]

RDF and OWL serialisations of SSSOM in the test data contain:

sssom:mapping_justification "semapv:LexicalMatching" ;

These should be:

sssom:mapping_justification semapv:LexicalMatching ;

This is not a bug with sssom-py, just with the test data. We should:

• [ ] Fix all the test data so that it looks like the latter.
• [ ] Add a test that fails if the generated rdf/owl serialisations does not turn EntityRef slots into proper URIs
• [ ] Add a test that fails if the generated data creates an <IRI> rather than a curie (can be achieved by testing file equality on string level rather than parsing and counting elements).