Use root-types to control input/output associations

[pkalita-lbl]

Fixes #29

When looking at GO-CAMs on NDEx, we noticed a lot of them had nodes (representing Activity instances) connected by both a direct regulation edge and a has input/output edge. See the linked issue for an example.

While we found the issue while looking at a rendering of the CX2 representation of GO-CAMs, these changes actually address the issue in the MinervaWrapper class where we convert from Minerva's JSON format into the gocam-py data model.

The logic here is to make more use of the root-types field in the Minerva format. In particular, these changes reflect a bit of pathway widget logic by checking that the object of a has input/output relationship has CHEBI:24431 (chemical entity) as a root type, but not CHEBI:33695 (information biomacromolecule) before adding a MoleculeAssociation. This is supported by tracking all of an individual's root-types (previously only one of them was tracked).

With these changes in place, no additional changes to the CX2 conversion process were needed. Here's an example of what the rendered CX2 for the example in #29 looks like with these changes after doing the full Minerva -> Model -> CX2:

[github-actions]

PR Preview Action v1.4.8
🚀 Deployed preview to https://geneontology.github.io/gocam-py/pr-preview/pr-34/
on branch gh-pages at 2024-12-05 01:07 UTC

[pkalita-lbl]

P.S. Sorry the diff is a bit messy because of my compulsion to replace variables like s with subject 😂