Algorithms for assessing integrity and quality of mappings

[cthoyt]

There are some great polar graph algorithms for this kind of stuff!

1. Look for n-cycles with all equivalent but one not-equivalent relation
2. Triangles and other cliques mean the same relation was curated many times. For each component, calculate the "density" and assign to all relations
3. Could namespace pairs have a prior confidence? Are there pairs of namespaces we expect don't go together, and therefore should have lower confidence in mappings?
4. Do more curation for a curator consensus (2 curators agree -> great!, 2 curators disagree -> not so much)
5. Look in connected components for the same namespace appearing twice