Refactor source retrieval provenance mdoel to better support edge merging use case

[mbrush]

The Architecture team has recently decided to pursue an edge merging strategy based on equivalent statement semantics + provenance (specifically, the identity of the original/primary source of the knowledge expressed in the edge). This new use case for “source retrieval provenance” has suggested potential shortcomings in the current standard for representing retrieval provenance.

The document here aims to clarify the nature/extent of these shortcomings, and offer proposals for enhanced representation of retrieval provenance that enable the expressiveness needed for the edge merging use case (and any other new use cases that have emerged).

When a specific proposal is accepted, tickets/PRs will be created in the TRAPI and Biolink repos to implement the required modeling.