Upgrade simulation API to PolicyEngine v4.10

[anth-volk]

Fixes #503

Summary

• Upgrade the simulation runtime to policyengine==4.10.0 while keeping existing API response contracts intact.
• Derive the deployed policyengine-us, policyengine-us-data, policyengine-uk, and policyengine-uk-data versions from the bundled policyengine.py release manifests instead of independently pinned/hard-coded values.
• Replace the US/UK-specific country package updater streams with one policyengine.py updater; that PR updates policyengine and syncs country package pins to the newly bundled versions in the same branch.
• Set gateway policyengine_bundle metadata from the same manifest-derived source, including policyengine_version, country model version, data package version, and manifest dataset URI.
• Define internal schema models for the single-year macro output and its budget, detailed budget, decile, poverty, inequality, geography, and labor-supply segments before dumping back to the existing dict contract.
• Centralize runtime output construction in src/modal/simulation_output_builder.py; SimulationOutputBuilder calls the PolicyEngine analysis/output functions, transforms raw .py objects into our segment schemas, and owns final build()/serialize() output assembly.
• Remove the separate simulation_output_adapter.py layer so the runtime builder is the single owner of .py calling, transformation, and serialization.
• Wire UK constituency and local authority outputs into the existing result keys and align dataset metadata with the PolicyEngine v4.10 bundle.
• Refresh the simulation project lockfile and strengthen focused contract/unit coverage.
• Add a model-agnostic AI guidance harness based on policyengine-us-data: canonical docs/engineering/skills/, thin tool-specific adapters, PR template, contributing guidance, and same-repository draft PR rules.

Testing

• uv run ruff check src/modal/simulation.py src/modal/simulation_macro_output.py src/modal/simulation_output_builder.py tests/test_simulation_output_builder.py fixtures/test_simulation_output_builder.py
• uv run pytest tests/test_simulation_output_builder.py tests/test_simulation_api_contracts.py tests/test_release_bundle.py tests/test_policyengine_dependency_source.py tests/test_modal_scripts.py tests/test_policyengine_package_update_scripts.py tests/gateway/test_endpoints.py tests/gateway/test_models.py -q
• git diff --check
• git diff --cached --check
• make -n push-pr-branch
• make -n help

[anth-volk]

Note for follow-up: this PR intentionally does not restore cliff_impact calculation yet. Before we wire that back into the simulation API, we need to add/restore cliff impact support in policyengine.py v4 as a reusable output surface, then have the API call that output when include_cliffs is requested. The old API behavior delegated this to .py via Simulation.calculate_economy_comparison(); the new v4 builder should not invent a separate API-local cliff implementation.