A locus tag is not read as a gene name, so a component covered by one reports as missing

[realmarcin]

Third review pass on #297.

gene_names() reads geneName and synonyms only:

value = (gene.get("geneName") or {}).get("value")
names += [s["value"] for s in gene.get("synonyms") or [] if s.get("value")]

UniProt also carries orderedLocusNames and orfNames, and several components
declare exactly those. attachment_organelle#hmw3_terminal_button declares
hmw3 and mpn452; the accession it already holds, Q50360, is gene hmw3 with
ordered locus MPN_452. So mpn452 is covered by the entry already on the
component, and --gaps reports it as NO_REVIEWED_ENTRY.

Four rows affected today: mpn309, mpn310, mpn452, and the same shape
elsewhere.

Why this was not fixed in #297

Making the report read locus names while the seeder does not would
recreate the report/seeder divergence #298 was filed for. Making both read them
changes what the seeder will match on, which is a real semantic change — a
locus tag identifies an entry, but matching on one is a different claim from
matching on a gene name, and it deserves its own canary and its own PR.

Excluding them may also be deliberate: gene_names() is the exact-match recheck
that stops the adapter trusting UniProt's looser search, and every name it accepts
widens that. Worth deciding explicitly rather than by extension.