Enforce PUF-clone weight-share floor in the dataset upload validator

[MaxGhenis]

Summary

The tax-unit construction engine and its rule helpers have been extracted into the standalone, reusable microunit package. This repository should depend on microunit for tax-unit construction rather than keeping its own in-repo copy, so the rules have a single canonical home and can be shared across PolicyEngine's microdata stack.

Current state

Tax-unit construction currently lives in:

• policyengine_us_data/datasets/cps/tax_unit_construction.py — the engine (construct_tax_units, modes, filing-status logic).
• policyengine_us_data/datasets/cps/tax_unit_rule_helpers.py — the federal rule helpers (CPS relationship codes, qualifying-child age test, qualifying-relative gross-income limit, etc.).

microunit was extracted from these modules: the engine logic (everything from the HEAD = "HEAD" marker onward) is byte-identical, and microunit exposes the same public API and the same construct_tax_units(person, year, mode="policyengine") -> (person_assignments, tax_unit) signature.

Proposed change

• Add microunit as a dependency.
• Delete the two in-repo modules.
• Re-point every call site (census_cps.py, the ACS wrapper, the two validation/ scripts, and the tests) to import from microunit.
• Keep policyengine_us_data/datasets/acs/acs_to_cps_columns.py and the thin ACS wrapper in this repo, since microunit deliberately excludes ACS column mapping.

This is behavior-preserving: tax-unit output is unchanged.

[MaxGhenis]

Closing as part of archiving this repository — Populace (PolicyEngine/populace) is its replacement and the certified US dataset since June 2026; see PolicyEngine/populace#204. This topic continues in populace#47 (PUF-clone weight-share floor enforcement; PR #1159). Published artifacts remain available but frozen.