Automate mapping of infores in primary_knowledge_source

[caufieldjh]

In discussion with @cmungall, it sounds like there is a demand for consistent use of primary and aggregator knowledge source in KGX output, particularly for ingest into larger KGs. Can we make this fully automated within KGX? The goal would be to populate primary_knowledge_source for edges based on a basic heuristic of inferring the source from the namespace, though this makes assumptions about whether edges are from imports or base files, etc. There are cases like Upheno and Uberon in which it's immediately clear what the primary source is (e.g., it's not HP, it's Upheno).

I see that some work towards this happened in #314. I have a hacky approach in kg-phenio too.

[caufieldjh]

Working on this here: #434
I think most of this can go in utils/infores.py since that's already comprehensive re: provided_by and knowledge_source,
but ideally this should be as automatic as possible, so if there's an edge between two CHEBI nodes and we haven't been provided with other provenance then the edge gets infores:chebi for primary_knowledge_source.