fix(control-plane): report mental model staleness accurately

[oldnicke]

Summary

• add an opt-in include_stale query parameter that computes scope-aware staleness for listed mental models
• use exact per-model staleness in the control-plane summary instead of comparing against the global consolidation timestamp
• request metadata-only results for the stats view and keep unknown staleness from being reported as stale
• regenerate OpenAPI and generated clients, and expose the option through maintained Python and TypeScript wrappers

Testing

• ./scripts/hooks/lint.sh
• uv run ty check hindsight_api/
• uv run pytest tests/test_reflections.py::TestMentalModelsAPI::test_mental_models_api_crud -v
• Python mental-model query mapping tests
• TypeScript mental-model query mapping tests
• go build ./...

Fixes #3139