Add operator runbook and redacted fixture guardrails for route review

[Disaster-Terminator]

Motivation

• Make the redacted route-review workflow usable and auditable without relying on raw prompts or private logs by providing operator guidance and machine-checkable fixtures.
• Enforce the contract that review cases must use expect as a product-level route_id (not target_model) so imports and audits compare stable route identifiers.
• Prevent accidental promotion of unredacted or target-model-based samples into eval banks by adding validation and a simple synthetic fixture for CI checks.

Description

• Add a concise operator runbook at docs/redacted_route_review_runbook.md describing how to collect redacted samples, required JSONL fields, import with route validation, how to run review_decisions against the decision endpoint, and how to interpret PASS/FAIL safely.
• Add a tiny synthetic redacted JSONL fixture at tests/samples/redacted_review_samples.synthetic.jsonl to serve as an operator sanity-check sample that contains redacted:true, text, and expect set to a route_id.
• Extend tests/test_import_review_samples.py with two tests that assert the fixture remains strictly redacted and non-empty and that convert_review_samples succeeds when validating against allowed route ids, and add Path import used by those tests.

Testing

• Ran the full test suite with uv run python -m pytest -q, which completed successfully with all tests passing (152 passed, 1 warning).
• Ran the eval smoke command uv run python scripts/eval_routes.py --mock-embeddings, which executed the mock eval cases and reported the expected passes (10 eval case(s) passed).

---

Codex Task

[sourcery-ai]

Hi @Disaster-Terminator! 👋

Your private repo does not have access to Sourcery.

Please upgrade to continue using Sourcery ✨

[qodo-code-review]

ⓘ You've reached your Qodo monthly free-tier limit. Reviews pause until next month — upgrade your plan to continue now, or link your paid account if you already have one.