Remove survey-based national calibration targets

[MaxGhenis]

Summary

• remove circular survey-derived/SPM-preservation targets from dense national calibration: SPM thresholds, capped housing subsidy, capped work-childcare, non-Part-B medical/OTC/premium aggregates, and gross alimony/child-support levels
• keep defensible component anchors for SPM-relevant expenses: ACS 2024 state/national contract rent and owner real-estate taxes, BLS CE 2024 childcare expenses, and IRS SOI itemized real-estate-tax amount/count targets
• add non-gross accounting constraints that alimony paid minus received and child support paid minus received equal zero, with absolute-error scaling so sparse ECPS keeps those zero targets
• keep age-bucketed health calibration limited to Medicare Part B and add regression tests for national loss/ETL target inclusion/exclusion
• add structured IRS SOI EITC-by-AGI-and-child-count target rows to policy_data.db and include that domain in local target selection
• add structured IRS SOI taxable-filer AGI/count targets by AGI band and filing status, including joint/surviving-spouse in constraints
• add policyengine_us_data.utils.national_target_parity, which maps every legacy national build_loss_matrix() label to a structured target ID or an explicit legacy/deprecated/lossy-label reason

Parity manifest

Generated from this branch against a rebuilt/temp policy_data.db and enhanced_cps_2024.h5:

• 921 national legacy loss labels
• 385 matched to structured target DB rows
• 536 classified as legacy-only, lossy legacy labels, or intentionally omitted with explicit reasons
• 0 accidental db_missing labels
• New taxable AGI/status constructor adds 148 DB rows; 136 legacy labels match exactly, while 12 old labels are explicitly classified as lossy because loss.py rounded away the true SOI AGI bucket boundaries.

Tests

• env -u UV_FROZEN uv run pytest tests/unit/calibration/test_loss_targets.py tests/unit/test_etl_national_targets.py tests/unit/datasets/test_enhanced_cps_seeding.py
• env -u UV_FROZEN uv run ruff format --check .
• env -u UV_FROZEN uv run towncrier check --compare-with upstream/main --staged
• git diff --check
• uv run ruff format --check .
• uv run ruff check policyengine_us_data/db/etl_irs_soi.py policyengine_us_data/utils/national_target_parity.py policyengine_us_data/calibration/calibration_utils.py policyengine_us_data/utils/constraint_validation.py policyengine_us_data/db/create_field_valid_values.py tests/unit/test_etl_irs_soi_overlay.py tests/unit/test_national_target_parity.py tests/unit/calibration/test_unified_matrix_builder_merge.py tests/unit/test_constraint_validation.py
• uv run pytest -q tests/unit/test_national_target_parity.py tests/unit/test_etl_irs_soi_overlay.py tests/unit/test_constraint_validation.py tests/unit/calibration/test_unified_matrix_builder_merge.py
• uv run pytest -q tests/integration/test_database_build.py
• uv run python -m policyengine_us_data.utils.national_target_parity --dataset-path /Users/maxghenis/PolicyEngine/policyengine-us-data/policyengine_us_data/storage/enhanced_cps_2024.h5 --target-db /tmp/policy_data_with_taxable_agi_status.db --period 2024 --output /tmp/policyengine-national-target-parity-with-taxable-agi-status.json