Adjust housing assistance take-up and validation

[MaxGhenis]

Summary

• increase the housing assistance take-up rate from 28% to 50%, with an inline note that 28% left ExtendedCPS modeled capped housing subsidies around half the raw CPS SPM benchmark
• lower the housing assistance formula/raw benchmark floor from 50% to 45% so the validation remains a missing-input guard rather than an exact calibration target
• add regression coverage for the observed production ratio, a clear-gap failure case, and the updated take-up parameter

Tests

• uv run pytest tests/unit/test_extended_cps.py -q
• uv run pytest tests/unit/test_take_up_parameters.py -q
• uv run ruff check policyengine_us_data/datasets/cps/extended_cps.py tests/unit/test_extended_cps.py
• uv run ruff check tests/unit/test_take_up_parameters.py