Add tests for improving RO-to-Biolink mappings

[gaurav]

Here are two ways of doing this:

1. We can look for cases where an RO predicate in an edge could not be directly mapped to a Biolink predicate. Generally this will be because it's too specific, and the more general mapping will be just fine, but making a list of these predicates could be useful in informing future improvements to the Biolink model.
2. Because cam-pipeline removes redundant triples, we should never see e.g. biolink:related_to, because we should always see a more specific predicate instead. So, a high-level predicate suggests that better predicates do not exist for the RO term.

These can probably be generated as debugging output from the Souffle scripts.