feat(proof): add evidence-backed claims ledger

[mrnicholasbcarter-code]

Summary

• add a versioned public proof matrix covering shipped contracts, observed evidence, partial gates, and explicit gaps
• add a claims ledger with allowed wording, provenance, confidence, falsification tests, downgrade wording, and review dates
• add redaction policy, adversarial review checklist, private-ledger template, and offline validator
• enforce proof-matrix validation in CI and document the public evidence entry point

Issue

Closes #134

Validation

• python scripts/verify_proof_matrix.py
• python -m pytest -q tests/test_proof_matrix.py tests/test_evidence_bundle.py tests/test_benchmarking.py tests/test_flagship_demo.py tests/test_verify_release_manifest.py
• ruff check scripts/verify_proof_matrix.py tests/test_proof_matrix.py
• ruff format --check scripts/verify_proof_matrix.py tests/test_proof_matrix.py
• mypy scripts/verify_proof_matrix.py --strict
• commit hook: Ruff, formatting, and mypy verdict --strict

The matrix intentionally marks unsupported quantified portfolio and production-readiness claims as downgraded until reproducible release evidence exists.

[github-actions]

📊 Benchmark Results

coding

Metric	Fixed	Verdict	Delta
Success Rate	100.0%	99.0%	-1.0%
Cost/req	$0.014488	$0.007163	50.6%
p95 Latency	146ms	145ms	-1ms

reasoning

Metric	Fixed	Verdict	Delta
Success Rate	100.0%	97.5%	-2.5%
Cost/req	$0.007650	$0.004623	39.6%
p95 Latency	147ms	144ms	-3ms

chat

Metric	Fixed	Verdict	Delta
Success Rate	96.0%	99.5%	3.5%
Cost/req	$0.000102	$0.001342	-1215.7%
p95 Latency	72ms	142ms	70ms