Fix Head Start categorical eligibility: split person/family aggregation

[hua7450]

Summary

Fixes a bug where Head Start categorical eligibility incorrectly aggregated person-level conditions (foster care, homelessness) across the entire tax unit. Under the old formula, if one child was in foster care, all children in the tax unit were shown as categorically eligible.

Changes:

• Splits categorical eligibility into two aggregation levels per 45 CFR 1302.12(c)(1):
  • Family-level (TANF, SSI, SNAP): aggregated at spm_unit — if the family is eligible for public assistance, all members qualify [(c)(1)(ii)]
  • Person-level (was_in_foster_care): not aggregated — only the specific child qualifies [(c)(1)(iv)]
  • Household-level (is_homeless): projected from household — all members in a homeless household qualify [(c)(1)(iii)]
• Switches aggregation entity from tax_unit to spm_unit (better represents "family" for benefit eligibility)
• Updates reference URL to headstart.gov and cites correct subsection (c), not (b)
• Fixes pre-existing reference tuple bug (missing comma between URLs)
• Restores categorical_eligibility.yaml with updated references including ACF-IM-HS-22-03 (SNAP as public assistance, April 2022)

Closes #7904

Regulatory basis

• 45 CFR 1302.12(c)(1): "A pregnant woman or a child is eligible if... (ii) The family is eligible for or... would be potentially eligible for public assistance; including TANF child-only payments"
• ACF-IM-HS-22-03 (April 2022): Expanded "public assistance" to include SNAP
• 2016 Final Rule preamble: Confirmed SSI as "public assistance" for Head Start

Test plan

• 30 Head Start tests pass (6 existing + 3 new edge cases)
• Foster care isolation: child in foster care does not make sibling categorically eligible
• Family aggregation: parent's SSI makes child categorically eligible via spm_unit
• TX household at 140% FPL: categorically eligible via computed SNAP
• CI passes

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (edaf7b1) to head (77d1ff4).
⚠️ Report is 53 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main     #7905      +/-   ##
===========================================
+ Coverage   97.26%   100.00%   +2.73%     
===========================================
  Files           5         1       -4     
  Lines          73        11      -62     
  Branches        2         0       -2     
===========================================
- Hits           71        11      -60     
+ Misses          1         0       -1     
+ Partials        1         0       -1     

Flag	Coverage Δ
unittests	100.00% <100.00%> (+2.73%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.