Add saved-H5 rescoring and household-metrics workflow

[MaxGhenis]

Summary

• add a saved-H5 rescoring workflow for rerunning reforms on locally built long-run calibrated datasets
• split scenario simulation from weighted aggregation via a household-metrics-first path
• add local parallel fanout tooling for year x scenario materialization
• track representative and all-reforms long-run findings from the saved-H5 checks
• document the methodology caveats around dependency behavior in the late tail

Scope

• extend src/year_runner.py with household-level scenario metrics and explicit weighted aggregation helpers
• add scripts/materialize_household_metrics.py
• add scripts/materialize_household_metrics_grid.py
• add scripts/aggregate_household_metrics.py
• update scripts/score_saved_h5_reforms.py, batch/compute_year.py, and modal_batch/compute.py to use the new aggregation path
• update README.md for the saved-H5 and household-metrics workflows
• add focused regression coverage in tests/test_year_runner_household_metrics.py

Method

• this PR is designed to work from locally built calibrated H5 datasets rather than waiting on policyengine-us-data CI
• it validates saved-H5 metadata before scoring and composes the intended baseline tax reform during rescoring
• baseline is treated as just another scenario in the new household-metrics path
• scenario simulation can now run ahead of final weighted aggregation, as long as the year support dataset exists

Validation

• uv run pytest tests/test_saved_h5_rescoring.py tests/test_year_runner_household_metrics.py -q
• uv run ruff check src/year_runner.py scripts/materialize_household_metrics.py scripts/materialize_household_metrics_grid.py scripts/aggregate_household_metrics.py scripts/score_saved_h5_reforms.py batch/compute_year.py modal_batch/compute.py tests/test_saved_h5_rescoring.py tests/test_year_runner_household_metrics.py
• python3 -m py_compile src/year_runner.py scripts/materialize_household_metrics.py scripts/materialize_household_metrics_grid.py scripts/aggregate_household_metrics.py scripts/score_saved_h5_reforms.py batch/compute_year.py modal_batch/compute.py

[github-actions]

Vercel site preview: https://crfb-tob-impacts-kp49tqbc5-policy-engine.vercel.app