Two different BacDive strains are both flagged as the type strain of one species

[realmarcin]

Confirmed by an adversarial review pass (survived 2 of 3 refuters) and re-verified on current main.

Defect

src/taxonmech/seed.py:515

if type_strain_ids & set(ccs):
    entry["is_type_strain"] = True

type_strain_ids is the union of every type-strain designation from every LPSN name on the record. A strain is flagged if it matches any single deposit in that union. A species has exactly one type strain, so a one-deposit match is too weak a rule.

Worked example

data/taxa/bacteria/salmonella_enterica.yaml flags two different strains, both against the same correct LPSN name (lpsn:784857, which lists 8 designations):

strain	designation	classified_as	matched designations
bacdive:5117	LT2	—	ATCC 43971, CCUG 42060, DSM 17058
bacdive:23943	110 (K 108), CNCT SK 108, …	NCBITaxon:59201	DSM 27656

Both render as type strain on the taxon page with nothing to distinguish them. The likely cause is that LPSN's designation list for a species includes deposits belonging to the type strains of its subspecies, and subtree gathering pulls those subspecies strains into the species record.

Why no gate catches it

tests/test_corpus_integrity.py::test_type_strains_match_a_nomenclature_designation restates the seeder's own derivation rule (does the strain match a designation?), so it cannot fail on this.

Suggested fix

Require the match to identify a single strain per name — e.g. flag only the strain matching the most designations of the record's correct name, and record the rest as type_strain_candidate with the matched deposits — plus a corpus test that no record flags two type strains for one correct name.

Requires a corpus rebuild. Sequence after #62.