Convert PBW-import script to use RDF directly instead of going through Neo4J

[tla]

For the sake of sanity, reproducibility, and transparency, the PBW graphimportSTAR script/module needs to be converted to write directly to RDF, instead of going through Neo4J and expecting a neosemantics export to happen.

This also gives us an opportunity to fix bugs e.g. #2, and implement the metadata schema devised as part of erc-releven/.github#1.

[tla]

Status report: tests are updated and pass against current-production and newly-generated data. I have manually checked that the only newly-created assertions on the existing datastore come from sources that were affected by the bug in #2 ; I only need to check that there aren't any assertions in the old data that were erroneously skipped in the new generation.