XY-859: Add end-to-end research-to-execution loop scenario tests

[yvette-carlisle]

Summary

• Add orchestrator runtime scenario coverage for latent research, promotion, Execution Program readiness, phase-goal validation, review checkpoints, guardrails, uncovered direction, and harness feedback.
• Add a research-to-execution loop operator runbook and runbook index entry.
• Refresh the test-suite inventory to the observed 1024 default-runnable nextest tests.

Dogfood assessment

• The architecture is meaningfully better than the old external research to manual Linear issue flow when the scenario evidence is present: latent research stays non-executable, promotion creates runtime-owned execution shape, only ready nodes get queued, phase-goal completion runs validation, review/validation churn has guardrails, and harness telemetry emits improvement recommendations.
• Remaining gaps: deterministic fixtures do not replace live Linear/GitHub writeback coverage for every path, native issue-generation ergonomics still need continued dogfood, and promotion remains an operator or runtime-policy decision.

Verification

• cargo make fmt
• cargo make lint-fix
• cargo test -p decodex --lib loop_scenario
• cargo make test (1024 tests run, 1024 passed, 1 skipped)
• semantic drift audit helper plus CLI help checks for research/status/evidence commands