Migrate poverty dashboard fixtures off split state H5 datasets

PolicyEngine/populace#237 removed release generation for one-file-per-state and one-file-per-district artifacts. The replacement path is the single certified Populace national H5 plus the standard region/scoping machinery in policyengine.py, not direct hf://policyengine/policyengine-us-data/states/... or districts/... files.

GitHub code search still finds split-H5 references in this repo:

• data/baseline.json
• tests/test_poverty_calc.py

Migration checklist:

• Stop constructing or documenting policyengine-us-data/states/*.h5 and policyengine-us-data/districts/*.h5 dataset URIs.
• Use the certified Populace default dataset through policyengine.py, with region filters/scoping for state and congressional-district analyses.
• Add or update a regression test so future state/CD runs do not depend on split regional H5 files.
• Keep historical/generated output unchanged unless it is actively used by the app or docs.

Start with data/baseline.json, tests/test_poverty_calc.py, and the standard region/scoping machinery in policyengine.py. Search these locations for split state and district H5 URIs, then run the poverty calculation tests. Done means state and congressional-district analyses use the certified Populace national H5 through policyengine.py, with regression coverage preventing split-file dependencies.