Don't do subclass_of stuff on a subclass_of query

[cbizon]

For instance, if you run this against ontological hierarchy:

{
    "message": {
        "query_graph": {
            "edges": {
                "e00": {
                    "subject": "n00",
                    "object": "n01",
                    "predicates": [
                        "biolink:subclass_of"
                    ]
                }
            },
            "nodes": {
                "n00": {
               
                    "categories": [
                        "biolink:Disease"
                    ]
                },
                "n01": {
                    "ids": [
                        "MONDO:0009242"
                    ],
                    "categories": [
                        "biolink:Disease"
                    ]
                }
            }
        }
    }
}

It doesn't really make sense to do the subclass_of for n01 in the cypher query. It causes there to be a bunch of crummy self edger results like:

{
        "node_bindings": {
          "n01": [
            {
              "id": "MONDO:0024543",
              "query_id": null,
              "attributes": null,
              "qnode_id": "MONDO:0009242"
            }
          ],
          "n00": [
            {
              "id": "MONDO:0024543",
              "query_id": null,
              "attributes": null
            }
          ]
        },
        "

[cbizon]

It's even worse when you are looking for superclasses of X. You start looking for subclasses of X and their superclasses. But there's no guarantee that a superclass of a subclass of X is a subclass of X. What gets returned is technically correct because of the qnode_id markings, but it leads to confusion, as well as an extreeeemly slow query.

See broadinstitute/molecular-data-provider#308

[EvanDietzMorris]

This issue is fixed in dev now. I could not reproduce the results from @vdancik's query (broadinstitute/molecular-data-provider#308), but using the example query here above shows that in dev (https://automat.renci.org/#/ubergraph/) we are no longer performing subclass inference for nodes that are connected to subclass_of or superclass_of edges.