Decide how US fiscal refresh should repair underweighted base H5 mass

[daphnehanse11]

Summary

The current published policyengine/populace-us 2024 H5 is not national-scale. Loading the latest cached artifact through the release builder gives 56,918,517 resolved weighted people, while the demographics benchmark used by Populace is 334,200,000 people (relative error -83.0%).

Because tools/build_us_fiscal_refresh_release.py calibrates with mass='conserve', the fiscal refresh cannot scale that base up to national fiscal/count targets. This explains the broad post-calibration failures we saw on #67 and the ACA support clipping noted on #74: the source refresh can assign every eligible support record and still cannot reach national enrollment/APTC targets from a one-sixth-scale base.

PR #93 adds a fail-fast guard so we do not spend a long calibration run or publish a release from this base silently. This issue tracks the actual repair decision.

Evidence

• Latest published release demographics reported total_population about 56.9M versus benchmark_total_population 334.2M.
• Local H5 inspection found household_weight sum 25.4M and resolved weighted person population 56.9M.
• ACA source diagnostics found 19.7M APTC-recipient target support but only 4.49M weighted eligible/APTC-support people in the base, causing take-up rates to clip at 1.0.
• A current fiscal refresh run then failed critical national targets by roughly 75-86%, consistent with conserved mass from a one-sixth-scale base.

Decision Needed

Pick and document one path:

1. Publish or select a national-scale base H5 before fiscal refresh.
2. Let the fiscal refresh calibration use mass='free' for this release path.
3. Explicitly rescale base weights to a national population control total before calibration, with a MassChange record and manifest evidence.
4. Some other source-stage fix that restores national person mass before calibration.

Acceptance Criteria

• The US fiscal refresh starts from a base whose resolved person population is near national scale, or intentionally changes mass with a documented mechanism.
• The release manifest records the base population scale / mass decision.
• A 2024 fiscal refresh can proceed past the base-population gate and reach the critical target gate without the broad conserved-mass failures seen on US fiscal-target build (0c38bc4): income tax regressed ~15% and SS dependents overshoot ~60% vs prior build #67.
• ACA enrollment/APTC diagnostics are rerun after the mass fix to separate true healthcare support gaps from denominator issues.

Related: #67, #74, #75, #93.