Expand OBBBA reform validation components

[PavelMakarchuk]

Summary

• expand the shipped OBBBA reform validation config from 2 rows to 19 model-backed components
• add FY2026 JCX-35-25 benchmarks for the identifiable JCT rows
• allow out-of-sample reform specs to carry a null JCT score for modeled rows without a standalone benchmark, currently the temporary senior deduction
• update reform-validation tests so OBBBA rows may be costs, revenue raisers, or scoreless modeled diagnostics

Notes

• The rows are counterfactual reverts because OBBBA is already in the policyengine-us baseline. The published effect remains baseline minus reform so signs are comparable to JCT enactment scores.
• I did not add unmodeled rows such as Trump accounts, employer-provided child care credit, dependent-care assistance exclusion, adoption credit changes, or corporate-only charitable floors as runnable validation rows.

Validation

• jq '.reforms | length' packages/populace-build/src/populace/build/us/obbba_reforms.json
• uv run --python 3.13 --package populace-build pytest packages/populace-build/tests/test_reform_validation.py
• uv run --python 3.13 --extra us --package populace-build python - <<'PY'
  from populace.build.us.reform_validation import out_of_sample_reform_specs
  specs = out_of_sample_reform_specs(period=2026)
  for spec in specs:
  spec.build_reform()
  print(f'built {len(specs)} reforms')
  PY
• uv run --python 3.13 ruff check packages/populace-build/src/populace/build/us/reform_validation.py packages/populace-build/tests/test_reform_validation.py