Closes #1257

Drug Target Commons builds heavily on the subset of ChEBML that is about drug-like or clincally tested molecules.

Questions

• Do we want to keep chemicals that are not named? We've had a few discussions about this already, and I think the consensus is no.
• Do we want to keep InChI keys as db_refs?
• How much processor-side processing should we do for targets that aren't given with uniprot identifiers, but either as a name that looks like a uniprot id, a name that looks like a gene symbol, a name that looks like a long name, or a name that looks like a comma-separated list of gene/family names? @bgyori I'm sure that code for parsing a field of mixed stuff exists already, and I think it would be good to reuse that if possible. Even if some refactoring/generalization is necessary, this would be another good addition to the indra.ontology.standardize submodule