Migrate impact computation off split state H5 files

Populace US releases are moving to a single national populace_us_2024.h5 plus region scoping, so this repo should stop downloading policyengine-us-data/states/{STATE}.h5 in scripts/compute_impacts.py.

Current references:

• scripts/compute_impacts.py:get_state_dataset() downloads states/{STATE}.h5 from policyengine/policyengine-us-data.
• .claude/commands/score-bill.md documents the same state-specific dataset path.

Expected migration:

• Use the certified national Populace dataset and scope by state_fips / congressional_district_geoid through policyengine.py region scoping, or a shared helper that preserves equivalent semantics.
• Do not synthesize one-H5-per-state or one-H5-per-district URLs.
• Keep district calculations based on the national file's congressional_district_geoid support.

Note: the current local checkout has active uncommitted work on add-manifold-markets, so I did not patch it in place.

Start with get_state_dataset() in scripts/compute_impacts.py and compare its dataset selection with the path documented in .claude/commands/score-bill.md. Trace how policyengine.py applies region scoping, then migrate both references to the certified national dataset while preserving state_fips and congressional_district_geoid calculations. Done means no state- or district-specific H5 URLs remain and district calculations use the national file.