The 26 groundable components have no UniProt accession to match on

[realmarcin]

Split out of #174 — the prerequisite for the half of it that InterPro can
actually close.

26 of the 49 ungrounded components are PROTEIN, so InterPro applies. None of
the 26 has a protein_examples accession. I checked all of them.

interpro_groundings.py matches on UniProt accessions, not gene symbols:

API = "https://www.ebi.ac.uk/interpro/api/entry/interpro/protein/uniprot"

So the 35 components carrying gene_symbols are not, as #174 originally
implied, ready to be matched. A gene symbol is not an identifier — flgE alone
does not say which organism or which entry.

Fix

scripts/uniprot_sl.py proteins already seeds protein_examples with reviewed
accessions where the record's GO term has a UniProt subcellular-location
mapping, restricted to a taxon the record names. That has been run for the
carboxysome and magnetosome. Running it across the other records is the
mechanical part of this.

Where no SL mapping exists, an accession can still be found the way the flagellum
ones were:

gene:<symbol> AND taxonomy_id:<taxon named on the record> AND reviewed:true

which is a query, not a guess — it returns a reviewed entry or nothing.

Order

This blocks the InterPro half of #174 and should be done first. #182 (complexes)
is independent and can proceed in parallel.