feat: add work order failure-code categorization scenarios

[ShuxinLin]

Summary

• Add src/scenarios/local/workorder_utterance.json with 6 focused WorkOrder-type evaluation scenarios for failure-code categorization (IDs 401–406).
• Scenarios exercise the wo/ server's get_failure_codes and get_work_order_distribution tools across three task categories:
  • Knowledge Query — list failure-code categories / codes under a category
  • Categorization — map a primary code, a WO description, or distinguish preventive vs. corrective by category
  • Distribution Analysis — group an asset's (CWC04013) work orders by failure-code category
• Add an evaluation loader test (test_workorder_scenarios_load_and_conform) verifying the file parses via load_scenarios and conforms to the Scenario schema.

This is the first workorder scenario set under src/scenarios/local/ (previously only vibration_utterance.json). Uses the PascalCase type convention ("WorkOrder") consistent with the sibling vibration scenarios.

Test plan

• uv run pytest src/evaluation/ -v -k "not integration" — 41 passed, including the new loader test
• uv run pytest src/ -q -k "not integration" — 313 passed; the lone failure (iot ... test_unknown_asset) is pre-existing on main and requires a local CouchDB (unrelated to this change)
• New scenario JSON validates against the Scenario model

Closes #350