Record strike_benefits as a reviewed degenerate input

[MaxGhenis]

strike_benefits joins US_DEGENERATE_INPUT_REVIEWED_EXCLUSIONS in tools/build_us_fiscal_refresh_release.py.

Why

Route A's fresh-base release run on 2026-09-23 (epic #956) refused at the pre-solve degenerate-input gate:

strike_benefits: every observed value equals the engine default (0)

alimony.py splits ASEC other income on OI_OFF: code 20 is alimony, code 12 is strike benefits, and everything else is miscellaneous income. policyengine-us-data's cps.py uses the same codes. The ASEC 2024 public use data dictionary confirms that OI_OFF 12 is "strike benefits", so the mapping is right. No respondent reports it. OI_OFF values in the three pinned official person files:

Person file (income year)	Codes present	Code 12 rows	Code 20 (alimony) rows
pppub23 (2022)	0, 2-10, 13, 19, 20	0	88
pppub24 (2023)	0-5, 7-10, 13, 19, 20	0	76
pppub25 (2024)	0-4, 7, 9, 13, 19, 20	0	75

The PUF half carries no strike-benefit field. So the column is identically zero on the pinned sources, and the gate's documented remedy is a reviewed exclusion, not a code change.

Guard rails

• The register's own rule still applies: if a later vintage has a code-12 reporter, the column stops being degenerate, the entry goes stale and the gate fails again.
• test_strike_benefits_exclusion_rests_on_zero_code_12_reporters reads the three official archives when MICROCOSM_ASEC_ARCHIVE_DIR points at them. It asserts there are no code-12 rows and that the alimony code is still present, which catches a layout change. It passes locally against the pinned archives (sha256 d2e00025…, cdb39cda…, 318845a2…).

Route A builds from route-a-integration-20260923, which merges this branch with #991 and #992.

🤖 Generated with Claude Code