[Thermo-Nuclear][Wave 4] Decompose oversized test suites (harness & evidence) below 1,000 lines #649

Parent #645 (Program Epic). Detail report: docs/audit/thermo-nuclear-636/04_codebase_file_size_and_complexity_thresholds.md.

Problem

Multiple test files exceed the 1,000-line ceiling:

• tests/test_verify_harness.py: 1,437 lines.
• tests/test_evidence_durability.py: 1,016 lines.
• tests/test_evidence_settlement.py: 1,003 lines.

Scope

• Decompose tests/test_verify_harness.py into:
  • tests/test_verify_harness_archive.py (~350 lines): intent archive chain integrity and durable specs.
  • tests/test_verify_harness_playbooks.py (~450 lines): playbook checks, parity, autopilot, and poteto-mode tests.
  • tests/test_verify_harness.py (~630 lines): core harness assertions.
• Extract shared mock fixtures and receipt generators into tests/evidence_test_helpers.py (~250 lines), bringing both test_evidence_durability.py and test_evidence_settlement.py well under 800 lines.
• Update scripts/select_checks.py check mappings and verify all tests pass without regression.