acetyl_phosphate --produces--> ATP has a CHEMICAL subject that neither produces nor derives-into fits well

[realmarcin]

Examined while resolving #327 and deliberately left alone; filed so the judgement is on record.

data/traits/metabolism/acetogenesis.yaml carries:

- subject: acetyl_phosphate     # node_type: CHEMICAL
  predicate: produces
  object: atp
  description: The acetate kinase step generates ATP via substrate-level phosphorylation.
  predicate_id: METPO:2007800

It is the last produces edge with a CHEMICAL subject, and it sits awkwardly
between two relations:

• METPO:2007800 (produces) — defined as "the activity or state of the subject
  brings the object into existence". A chemical has neither in the agentive sense; that is
  the same imprecision that moved the two thiosulfate edges in Two opposite-direction predicate labels share METPO:2000013, so one asserts the inverse of the term's meaning #327.
• RO:0001001 (derives into) — requires that the new entity "inherits the significant
  portion of the matter of the old entity". This one fails that test. ATP's matter comes
  mostly from ADP; only the phosphoryl group transfers from acetyl phosphate. So it is not
  a substrate-to-product transformation in the sense the other three were.

Because it fails derives into on the ontology's own wording, it was not swept into #327
alongside the thiosulfate pair. It keeps METPO:2007800, which is why that row's
subject_types still admits CHEMICAL.

Options

1. Re-point to the process. The real producer is the acetate kinase step, already a
   node in that graph. <acetate_kinase_step> --has output--> atp (RO:0002234) is
   domain-correct and true. The phosphoryl-donor role of acetyl phosphate would then want
   its own edge to survive.
2. A phosphoryl-transfer / group-donor relation, which is what the edge is really
   asserting. No obvious upstream term; would mean a METPO proposal for one edge.
3. Accept and document — the current state, with the reasoning recorded here and in the
   mapping row.

Option 1 looks strongest and is the only one that lets METPO:2007800's subject_types
drop CHEMICAL entirely, which would make that gate exactly match its definition.

Related: #327 (the three edges that did move), #330 (the other predicate whose corpus usage
does not match its term's sense).