Update agent docs away from split state/district H5 datasets

[MaxGhenis]

PolicyEngine/microcosm#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:

• skills/tools-and-apis/policyengine-microsimulation-skill/SKILL.md
• skills/analysis/policyengine-district-analysis-skill/SKILL.md
• skills/domain-knowledge/policyengine-healthcare-skill/SKILL.md

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.