Add Medicaid and CHIP eligibility-to-enrollment diagnostics

[daphnehanse11]

Context

The current published Populace US 2024 dataset shows Medicaid and CHIP enrollment equal to eligibility in aggregate:

• is_medicaid_eligible: 72.30M; medicaid_enrolled: 72.30M.
• is_chip_eligible: 4.98M; chip_enrolled: 4.98M.
• Medicaid + CHIP enrolled: 77.28M vs CMS combined target 79.15M (-2.4%).

This means Medicaid/CHIP take-up is effectively 100% among modeled eligibles. That may reflect the current PE-US variable semantics, but Populace release diagnostics should make the eligibility-to-enrollment assumption visible rather than implicit.

This complements #139/#165, which add direct CHIP targets, but it is a separate input-signal/diagnostics question.

Goal

Add Medicaid and CHIP eligibility-to-enrollment diagnostics, then decide whether Populace should materialize enrollment/take-up inputs or leave the behavior formula-owned.

Suggested implementation

• Add release diagnostics for Medicaid and CHIP:
  • eligible count,
  • enrolled count,
  • enrollment / eligibility ratio,
  • Medicaid + CHIP combined count,
  • same breakdown by state where source support exists.
• Add a gate or warning when modeled enrollment is mechanically identical to eligibility for every state/program.
• Document whether medicaid_enrolled and chip_enrolled are Populace inputs, PE-US formula outputs, or a mixed source/formula boundary.

Acceptance criteria

• Diagnostics make 100% eligibility-to-enrollment take-up visible nationally and by state.
• Tests cover nondegenerate and degenerate take-up cases.
• If the variables are formula-owned by PE-US, Populace records that boundary and links a model-side issue.
• If Populace owns source inputs, a follow-up issue/PR scopes Medicaid/CHIP take-up materialization separately from direct target mapping.