Replace flat voluntary filing rate with demographic elective-filer model

[MaxGhenis]

Summary

• replace the flat voluntary filing scalar with a 16-cell demographic table over (children bin, wage-income bin, age-head bin)
• switch the voluntary-filing pool from raw takes_up_eitc propensity to actual modeled EITC claimants, so non-claimants remain eligible for elective filing
• add a filer-alignment metric to sanity_checks.py and focused unit tests for the new lookup helpers

Calibration shape

• num_children: no_children, with_children
• wage_income: zero, low ($0-$15k), medium ($15k-$30k), high ($30k+)
• age_head: under_65, age_65_plus

Validation

• uv run pytest tests/unit/test_stochastic_variables.py tests/unit/datasets/test_cps_voluntary_filing.py -q
• uv run pytest tests/unit/datasets -q
• uv run python policyengine_us_data/calibration/sanity_checks.py policyengine_us_data/storage/enhanced_cps_2024.h5 --period 2024
• patched-temp H5 sanity check improved non_eitc_filer_total_gap from -9.9M to -3.9M, non_eitc_filer_total_agi_abs_error from 11.40M to 10.97M, and non_eitc_filer_low_agi_abs_error from 3.33M to 2.90M

Closes #813