Add replay semantic integrity artifact

[ProfRandom92]

Summary:
Add a deterministic replay semantic integrity artifact and generator that summarizes operational commitment-class survival across all manifest-registered fixture families, relying only on manifest inputs, existing admissibility artifacts, and registered failure labels.

Changed files:

• scripts/generate_replay_semantic_integrity_artifact.py
• artifacts/replay_semantic_integrity_results.json
• tests/test_replay_semantic_integrity_artifact.py
• package.json (added generate:replay-semantic-integrity script)

Testing:

• Ran targeted pytest: pytest tests/test_replay_semantic_integrity_artifact.py -q (7 passed).
• Verified related artifact tests: pytest tests/test_mcp_trace_replay_artifact.py -q (5 passed) and pytest tests/test_multi_family_admissibility_artifact.py -q (6 passed).
• Verified manifest and taxonomy tests: pytest tests/test_fixture_manifest.py -q (8 passed) and pytest tests/test_failure_taxonomy.py -q (4 passed).
• Ran full validation: npm run check which executed the project test suite and build, resulting in all tests passing (234 passed in 50.74s).

Risks:

• Commitment-class mapping is a small deterministic keyword/layer mapping and may classify generic or ambiguous contract IDs into the fallback constraints class, which is test-covered but may require fine-tuning later.

Next:

• Open PR (branch feature/replay-semantic-integrity-artifact) and monitor CI artifact drift/reproducibility gates to ensure the committed artifact remains stable in remote CI.

---

Codex Task

[gemini-code-assist]

Code Review

This pull request introduces a new artifact generation pipeline for replay semantic integrity results, consisting of a generation script, a new JSON artifact, and a comprehensive test suite. The code reviewer identified opportunities to modernize and optimize the Python script by removing redundant OrderedDict usage and streamlining the data aggregation logic for better performance.

[gemini-code-assist]

The OrderedDict import is unnecessary as standard dictionaries in Python 3.7+ preserve insertion order. Using regular dictionaries simplifies the code and reduces overhead.

[gemini-code-assist]

The aggregation logic can be optimized by avoiding list concatenation and using set.update() for failure labels. Additionally, OrderedDict is redundant here as standard dictionaries maintain insertion order.

        commitment_classes = {
            cls: {"passed": 0, "failed": 0, "failure_labels": set()}
            for cls in COMMITMENT_CLASS_ORDER
        }

        for point in points:
            failed_contracts = set(point["failed_contracts"])
            for contract_id in point["passed_contracts"]:
                commitment_classes[_class_for_contract(contract_id)]["passed"] += 1
            for contract_id in point["failed_contracts"]:
                cls = _class_for_contract(contract_id)
                commitment_classes[cls]["failed"] += 1
                commitment_classes[cls]["failure_labels"].update(point["failure_labels"])

        serializable_classes = {}
        for commitment_class in COMMITMENT_CLASS_ORDER:
            values = commitment_classes[commitment_class]
            serializable_classes[commitment_class] = {
                "passed": values["passed"],
                "failed": values["failed"],
                "failure_labels": sorted(values["failure_labels"]),
            }