Populate remaining US tax-input and reported-observation layers

Summary

Track the remaining US input-layer work as source/profile work in live Populace, independent of any incumbent comparison surface.

Inputs to populate or explicitly classify

Pure tax inputs that should be sourced from the PUF donor/impute stage:

• amt_foreign_tax_credit
• general_business_credit
• excess_withheld_payroll_tax
• prior_year_minimum_tax_credit
• early_withdrawal_penalty
• other_credits

Formula-owned output with missing upstream support:

• self_employed_health_insurance_ald should not be persisted directly. Populate the upstream self-employed health-premium inputs needed for the PolicyEngine-US formula.

Reported/comparison observations to classify:

• spm_unit_net_income_reported
• spm_unit_total_income_reported
• ssi_reported

These are not formula-consumed benefits/taxes; decide whether Populace should carry them as reported observations or document their absence as named exclusions.

Acceptance criteria

• The source/profile inventory records each variable as populated, upstream-populated, carried as reported observation, or reviewed-excluded.
• The writer still rejects formula-owned outputs as final inputs.
• Tests cover at least one populated PUF-derived credit, the SE-health upstream path, and the reported-observation classification.
• Release diagnostics expose any reviewed exclusions.

Supersedes live Populace issue #26, which framed this as an incumbent-comparison gap. Benchmark/incumbent comparison discussion now belongs in PolicyEngine/populace-benchmarks.

Start with the source/profile inventory and writer behavior for the listed tax and reported-observation variables. Trace the upstream self-employed health-premium path and run the existing tests before adding coverage for a PUF-derived credit, that path, and observation classification. Done means every variable has a recorded classification, formula-owned outputs are rejected as final inputs, and release diagnostics expose reviewed exclusions.