Wire the confidence facets together

[edkerk]

The three per-reaction confidence scorers (localization, equation, gene_association) shipped as separate calls with no glue. This wires them into a pipeline and closes the loop between confidence and curation — the "wire the facets together" work item.

annotate_confidence — one call, every applicable facet

annotate_confidence(model, proposal=..., scores=...) runs every scorer and returns {facet: count}. equation and gene_association need only the model; localization runs only when a proposal and its scores are supplied, and is skipped rather than failed otherwise — the same abstain-rather-than-guess rule the scores themselves follow. The returned dict reports what actually ran, and facets=[...] restricts the set (an unknown facet raises).

curation_priority reads the record — the loop closes

A placement a curator has settled with mark_curated now drops out of the review queue (new respect_curated=True, default on). So the pipeline is a closed loop:

score → review here → curate (mark_curated) → stop being asked

The skip is targeted (only that reaction's placement rows — a second flagged reaction stays) and reversible (respect_curated=False ranks everything as before). Transports and gap-fills are unaffected, since curation settles a reaction's placement, not the bridges around it. The confidence import is lazy, to avoid the localization ↔ confidence cycle.

Smaller fix

The no-SBO warning now names its remedy — raven_toolbox.annotation.add_sbo_terms(model) — instead of just saying "annotate SBO terms first".

Tests & verification

New tests cover the umbrella (facet selection, localization skipped without a proposal, unknown-facet error) and the curated skip (flagged → suppressed → restored; targeted not global). pytest 842 passed / 4 skipped; ruff, mypy, and the -W docs build clean.

Note on the study doc

The confidence_tracking.md §10 plan that lists these as the next items is being rewritten in #73 (which removes the rejected reversibility facet). To avoid a merge conflict I left the doc to #73; once both land, its §10 should mark these three items done. Flagging so it doesn't fall through the cracks.

[edkerk]

Duplicate of #78, which is the canonical facet-wiring PR. Closing in favour of #78.