household_state_benefits omits 28 of 51 implemented state child care subsidy variables (child care assistance never reaches net income in those states)

## Summary

`gov.hhs.ccdf.child_care_subsidy_programs` lists 51 state/DC child care subsidy variables, but `gov.household.household_state_benefits` (the list that `household_state_benefits` → `household_benefits` → `household_net_income` sum) includes only 23 of them in every period block, including 2026. In the other 28 states an implemented child care subsidy never reaches household benefits or net income, so the household app, the marriage calculator, and any microsimulation that reads `household_net_income` show zero child care assistance there.

Follow-up to #9234 (closed) and #9236, which added NC, DC and NY. Verified on `main` at 1ca1d4ee (2026-09-07) by parsing both YAML files.

## Present in `household_state_benefits` (23)

al ar az ca co dc de ga id in md mo mt nc ne nv ny oh ok sc va wa wv

## Implemented but missing from `household_state_benefits` (28)

ak ct fl hi ia il ks ky la ma me mi mn ms nd nh nj nm or pa ri sd tn tx ut vt wi wy

## Per-period counts in `household_state_benefits.yaml`

| Block | entries | `*_child_care_subsidies` |
|---|---|---|
| 2023-01-01 | 54 | 16 |
| 2024-01-01 | 60 | 22 |
| 2025-01-01 | 61 | 23 |
| 2026-01-01 | 62 | 23 |

## Suggested fix

Either add the 28 variables to every applicable block of `household_state_benefits.yaml`, or replace the per-state entries with the single aggregate `child_care_subsidies` (which already `adds` `gov.hhs.ccdf.child_care_subsidy_programs`) and remove the 23 per-state entries to avoid double counting. The second option keeps the two lists from drifting again; a test that asserts every `child_care_subsidy_programs` entry is reachable from `household_benefits` would lock it in.

Context: needed for a PolicyEngine analysis of the HHS "parent-based child care" draft rule (Sept 2026), where CCDF amounts must flow into net income consistently across all states.

Start by comparing gov.hhs.ccdf.child_care_subsidy_programs with household_state_benefits.yaml, checking each dated period block and the existing child-care entries. Choose and implement a consistent aggregation approach, then add a regression test showing every implemented subsidy reaches household_benefits and household_net_income without double counting.

Written by the indexing model from the issue text.