Exempt NY and MN from per-state ACA calibration tests

[MaxGhenis]

Summary

• NY (Essential Plan) and MN (MinnesotaCare) are ACA §1331 Basic Health Programs — federal BHP funds cover 138–200% FPL residents who would otherwise claim APTC on the Marketplace.
• CMS's per-state APTC targets the calibration tests compare against reflect this diversion (NY target ~$0.86B, MN similar) but `policyengine-us` does not model BHP, so simulated `aca_ptc` stays at full-Marketplace levels.
• Result: every main-branch build since PR Impute selected_marketplace_plan_benchmark_ratio from CPS premiums #801 fails `test_aca_calibration` / `test_sparse_aca_calibration` on NY (511–880% vs 500% tolerance), blocking HF publish of new EnhancedCPS h5s — including the one that should include the EITC calibration work from Add state and AGI cross-tab EITC calibration targets (#802) #803.
• This PR exempts the two BHP states from the per-state spending tolerance check. All 49 non-BHP states continue to be checked at 500%. A log line still prints the BHP state error, tagged `[BHP, exempt]`.
• The structural fix (switch per-state target to IRS SOI total PTC claimed `A85770`, drop the CMS national anchor) is in Replace per-state CMS APTC target with IRS SOI Net Premium Tax Credit (A11560/N11560) #805. The upstream fix (model BHP in `policyengine-us` so `aca_ptc` zeros out for BHP enrollees) is separate.

Test plan

• `uv run ruff format --check` — clean
• CI `integration-tests` now pass on main
• Post-merge push.yaml completes build-and-test + publish, uploading the Add state and AGI cross-tab EITC calibration targets (#802) #803 EITC-calibrated EnhancedCPS to HF

🤖 Generated with Claude Code