Make sure subject is always in subject_closure, object is always in object_closure

[kevinschaper]

This is technically a closurizer issue, but I don't want it to get lost there, especially if we start doing our denormalization elsewhere.

Right now, monarch-py isn't finding associations (or association-counts) for genes, because it's only looking in subject_closure & object_closure. Those fields should be reflexive closures that include the self-ID along with he ancestor IDs, and when they're populated from relation graph files, they are...but for entities that aren't in RG, we need to manually populate the fields.

Over in monarch-py, we'll probably put logic in place to check both subject and subject_closure etc, but that's now how we want it to work.