omnicorp_overlay rarely works

[cbizon]

For instance, consider this strider query:

{
    "message": {
        "query_graph": {
            "nodes": {
                "n0": {
                    "id": "UniProtKB:P52788",
                    "category": "biolink:Gene"
                },
                "n1": {
                    "category": "biolink:ChemicalSubstance"
                }
            },
            "edges": {
                "e01": {
                    "subject": "n0",
                    "object": "n1"
                }
            }
        }
    }
}

This produces results in strider, and the message that arrives at omnicorp has a KG like:

...
'PUBCHEM.COMPOUND:68911': {'category': ['biolink:ChemicalSubstance'],
     'name': 'artemether'},
'NCBIGene:6611': {'category': ['biolink:Gene']}
...

At this point multiple things appear to go wrong.

1. Omnicorp is not returning counts for any nodes. This may be true for the chemical nodes (see below), but not for the gene. Directly consulting the postgres with NCBIGene:6611 returns many results
2. Curie prefixes are murder. For the pubchem.compound above, we don't have any results in the postgres, though we do for the related chebi. However, the KG arriving in omnicorp does not have equivalent_identfiers in it, so we can't troll through there for them. (See https://github.com/TranslatorSRI/NodeNormalization/issues/43)

Assuming that we can solve the NN issue above, and enable NN in the step prior to omnicorp, then we need to use the equivalent identifiers to check all the possibilities in omnicorp, and merge them together if there are results for multiple prefixes.

We also need to figure out why NCBIGene:6611 isn't returning anything, because it should.

[cbizon]

https://github.com/ranking-agent/strider/issues/164

[cbizon]

Needed rebuilt cache.