feat(evaluator): wire algebraic commutativity runner

[briansrls]

Summary

• add a bounded COMMUTATIVITY_WITNESS_PAIRS sample table alongside associativity witnesses
• wire AlgebraicLaw(Commutativity, lens_ref) through the runner using the existing predicate surface and a shared structural-value comparator
• keep Identity fail-closed with the missing identity-element edge called out, and keep distributivity routed to the P1 substrate enum-extension path
• add focused runner tests for commutativity pass/fail and the identity blocker

Verification

• cargo fmt --all --check
• git diff --check

Attempted focused test: cargo test -p v3-compiler test_runner_algebraic_law_commutativity --test integration.

• With the default target dir it failed with No space left on device while writing dependency files on the /Users worktree volume.
• Retried with CARGO_TARGET_DIR=/tmp/gunbc-pr-b3-target; dependency compilation proceeded but remained in v2-compiler compilation for about 30 minutes, so I stopped it rather than leave a runaway job.

[briansrls]

Review metadata

• Provider / model: cursor / composer-2
• Commit: d937f10a · Trigger: schedule
• Comparison: origin/main @ 79e61ba0 ... review/pr-1359-d937f10a @ d937f10a
• Thinking: 93s wall

---

Findings

• src/v3/compiler/tests/integration/test_runner_test.rs:1018-1021 (and similarly 1051-1054): New tests assert on human-readable ClaimResult strings via .contains(...). TESTING.md (“Anti-patterns” → don’t pin error message text) treats that as a brittle contract; for the subtraction case, matches!(..., ClaimResult::Fail(_)) (or a typed/stable carrier if one exists) would assert the behavior without coupling to wording.

Verdict

APPROVE_WITH_COMMENTS — The runner change mirrors the existing associativity pattern: bounded witness table in lens_apply.rs, compile-time len() > 1 ratchet, explicit “operational witness, not substrate proof” + dissolution language in test_runner.rs, and Identity stays fail-closed / NYI with a clear rationale. Nothing here looks like substrate duplication of authority beyond the already-documented runner scaffold. The only rubric-aligned nit is the new substring-based test assertions above.

Exploratory (optional)

• src/v3/compiler/src/test_runner.rs:305: The unreachable! after string matching is conventional for “validated twice” flow; if you ever want to avoid panic surfaces in this helper entirely, the inner match could return the same UnsupportedLaw/MalformedPayload path instead (implementation taste, not a modeling violation).

[briansrls]

Follow-up PR opened for the review nit about substring-based assertions: #1363

It replaces the message substring checks with typed ClaimResult::Fail(_) / ClaimResult::NotYetImplemented(_) matches. git diff --check and cargo fmt --all --check passed; focused test still stalls in dependency compilation in this environment, matching the original PR note.

— sent from snappy-moth-795