[codex] Mark E-P producer coverage gate landed

[briansrls]

Summary

• Updates the R3 substrate carrier roadmap row to record gate Lane B tasks #76 (e_p_per_call_descent_evidence_full_coverage) as landed for the current lowered callable surface.
• Refreshes the v3 lens capability register so remaining E-P work is scoped to cost-lens consumer/cementing and non-carrier blockers, not stale producer-coverage language.
• Updates the per_call_descent_evidence doc comment to reflect current per-argument coverage and fail-closed behavior.

Validation

• cargo fmt --all --check
• cargo test -p v3-compiler --test integration m2_substrate_inhabitance_test::e_p_per_call -- --nocapture
• cargo test -p v3-compiler --test integration complexity_lens_behavioral_completion -- --nocapture

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: 6aee38ac · Trigger: schedule
• Thinking: 46s wall

---

Non-blocking — Improvements (fix in-PR if easy, else defer to roadmap)

• src/v3/compiler/src/dag.rs 1267|This revised comment in per_call_descent_evidence claims broader coverage (match-payload/field-projection/nested/multi-argument classes), but the current producer logic still emits only PreservedValue, ArithmeticDescent, and SubValueUnknown, so this status text should stay aligned to the implemented slice or be updated with the missing producer logic.
• docs/v3-lens-capability-register.md 9|The register now states Gate #76 has already broadened E-P to the full live callable-callable classes, but that breadth is not yet reflected by the named e_p_* ratchets in m2_substrate_inhabitance_test, so this should be marked as partial with explicit lane/decomposition under T-E-P-Producer-Broadening.
• ROADMAP.md 408|ROADMAP flips the E-P lane to “producer coverage landed,” yet design/docs still describe T-E-P-Producer-Broadening as open with current first slice language; this is a sequencing/receipt mismatch and should stay as the follow-up lane until those additional ratchets and evidence are in-tree.

✅ No blocking issues found; the PR has non-blocking documentation/status-accuracy concerns only.

[briansrls]

Addressed the non-blocking status-accuracy feedback in c2145ba. I verified the producer does emit structural descents as SubValueRelation::StrictSubValue via match_payload_descent_relation / match_payload_field_projection_descent_relation, then tightened the rustdoc to name the emitted variants explicitly. I also updated r3-structure/ROADMAP/register language to scope gate #76 as the current ratcheted lowered-callable producer surface and leave broader lane work to consumer/cementing. — sent from warm-ram-852