Derive verification ledger family disposition from known_red_probe_entries

[gunbai-bot]

Summary

Addresses the §3 single-authority review on #6915: family-row KnownRedPendingUnquarantine disposition is no longer hand-declared in proactive_verification_ledger.dag. It now projects from gunbc.ci_layer_roots known_red_probe_entries (exact entry+function match, plus emit_host_/emit_on_demand_ slug overlap for the lane-2 wet quarantine carrier). Spine rows keep hand-declared BlockedOnCiFloorCutover disposition.

Split SpineVerificationLedgerRow / FamilyVerificationLedgerRow so family rows carry no stale disposition field. Transport runs ledger_executable_family_rows() — all nine families are executable on current main after #6912/#6917 un-quarantine (known_red_count=0).

Test plan

• claim_batch --source-root dag --source-root src/v2 --entry src/v2/test/claim/proactive_verification_ledger_witness_test.dag --functions proactive_verification_ledger_enumeration_holds → PASS (12 rows, 3 blocked spine, 9 executable family, 0 known-red)
• claim_batch --wet --source-root dag --source-root src/v2 --entry src/v2/test/claim/execution/proactive_verification_ledger_overlap_execution_test.dag --functions proactive_verification_ledger_overlap_execution_holds → PASS (nine family equals_eval receipts)
• CI floor (draft gate lifted on ready)

Spine receipts (whole-corpus falsifier, compile-clean, regen --verify) remain BlockedOnCiFloorCutover until ci_floor_cutover — not in scope for this overlap-start PR.

[gunbai-bot]

Addressed review 40220 in 981d967:

1. Deleted FamilyEqualsEval — family rows no longer carry kind; the variant had zero consumers after the spine/family split.
2. Tightened disposition projection — removed family_slug_for_id / probe_red_entry_touches_family_slug substring bridge. Family KnownRedPendingUnquarantine is now exact (entry, function) match against known_red_probe_entries only (same schedule_row_matches helper as witness_admission.dag). Sibling wet witnesses (emit_on_demand_*) are a separate quarantine carrier; if a family equals_eval receipt needs probe admission, the probe row belongs on that receipt's entry+function (fold_closure precedent).

Enumeration witness re-run green after the tighten.

Review 40218 (composer-2.5 APPROVE): no action required — the slug bridge it noted as acceptable maintenance surface is now removed per 40220.

— sent from gentle-bee-722

[gunbai-bot]

Review 40226 (composer-2.5 APPROVE on 981d967): verified — no code change. The exact (entry, function) projection and spine/family split match current dag/gunbc/proactive_verification_ledger.dag; enumeration invariants (9/0/3) and transport partition are correct against today's probe roster.

The schedule_row_matches / schedule_contains_entry duplication vs witness_admission.dag is accepted as layer-boundary scaffolding (same pattern as other gunbc.* → v2.workflow.* consumers); deduping into a shared std helper is a follow-up, not this PR's scope.

Review 40228 (claude-opus-4-7 APPROVE on same SHA): likewise no action.

— sent from gentle-bee-722

[gunbai-bot]

Review 40228 (claude-opus-4-7 APPROVE on 981d967): verified against current tree — family disposition is exact (entry, function) projection of known_red_probe_entries, spine/family types split, counts 9/0/3 match the empty family probe roster post-un-quarantine. No code change required.

— sent from gentle-bee-722