[codex] strengthen dogfood evidence validation

[baskduf]

What changed

• Added a dogfood evidence adoption checklist and linked it from evaluation, validation, and the component map.
• Extended effectiveness-plan validation to reject stale aggregate completion language, countable templates, placeholder task outcomes, and inconsistent inclusion flags.
• Synced the generic template checker and added unit coverage plus dogfood failure/decision memory.

Checks

• python3 -m unittest discover -s tests
• python3 -m py_compile scripts/apply_harness.py scripts/check_docs_drift.py scripts/check_structure.py scripts/check_encoding_hygiene.py scripts/check_effectiveness_plan.py scripts/check_failure_memory.py scripts/check_decision_memory.py scripts/harness_doctor.py
• python3 scripts/check_docs_drift.py
• python3 scripts/check_structure.py
• python3 scripts/check_encoding_hygiene.py
• python3 scripts/check_effectiveness_plan.py
• python3 scripts/check_failure_memory.py
• python3 scripts/check_decision_memory.py
• python3 scripts/harness_doctor.py --target .
• cmp -s scripts/check_effectiveness_plan.py templates/generic/scripts/check_effectiveness_plan.py

Review

Direct review and sub-agent review completed before commit; latest sub-agent review returned no findings.

Remaining risk

Harnessed-only dogfood evidence still does not prove effectiveness improvement without a comparable baseline or later comparison window.