16 PROTEIN_COMPLEX components have no grounding authority

[realmarcin]

Split out of #174.

interpro_groundings.py refuses non-PROTEIN components by design, and it is
right to: an InterPro family denotes a protein, and the MS ring is ~26 copies of
FliF. Tagging the ring with the FliF family would assert the ring is a protein.

InterPro grounding import refused: automatic InterPro review accepts only
PROTEIN components; GO:0009288#ms_ring is 'PROTEIN_COMPLEX'

That leaves 16 PROTEIN_COMPLEX components with no route to a grounding at
all — the flagellum's MS ring, C ring, stator, rod, LP rings and export
apparatus among them.

Complex Portal is the obvious candidate

It is already ADOPTED in the source queue and scripts/complex_portal.py
exists. A Complex Portal accession denotes exactly a defined multi-protein
complex, which is what these components are.

Worth checking first whether the flagellar sub-complexes actually have Complex
Portal entries — several may not, in which case a GO cellular-component term is
the fallback (GO:0009431 bacterial-type flagellum basal body, and similar).

A related modelling question

lp_rings combines FlgH and FlgI, and each has an exact InterPro family of its
own — IPR000527 (L-ring) and IPR001782 (P-ring) — while the combined
component has none. It is the one case in the flagellum where splitting the
component would make both halves groundable. Whether that is worth doing, or
whether the L and P rings are better kept as one component, is a curation
decision rather than a tooling one.