Processor updates to improve grounding-based preassembly

[bgyori]

• The BioPAX and TRIPS processors currently produce lists of HGNC and UP IDs. We need to prioritize within the processors so that internal INDRA processes get single values. Alternatively, we need to figure out how to handle lists of possible groundings in a useful way within INDRA.
• The BEL processor needs to map to INDRA families from the BEL family name space.
• The TRIPS processor represents ad-hoc protein families in db_refs in a way that currently crashes preassembly
• The TRIPS and REACH processors use FA (PFAM) and IP (InterPro) grounding. These should somehow align against the entity hierarchy.
• The BioPAX processor currently does a very expensive combinatorial expansion of ad-hoc families into their members. We should consider mapping these to INDRA families, although that would make them less specific than they are.

[bgyori]

• This is a true legacy BELpy "feature": processors should not standardize Agent names according to PySB standards

[bgyori]

• All processors should use str instead of int for all db_refs entries including HGNC IDs

[bgyori]

• Processors need to produce db_refs according to the newly defined patterns (e..g. "CHEBI:...")

[bgyori]

• Processors should differentiate between different entity types using the same name space (e.g. TRIPS using UniProt IDs for non-proteins).

[bgyori]

Done as of #141