Backfill IRS SOI source packages to 2015 for Microplex PUF uprating

Context

microplex-us currently needs a historical PE-style SOI target surface for PUF uprating, with values back to 2015 and forward through the PE-SOI/factor boundary used by the rebuild pipeline.

In PolicyEngine/microplex-us#155, the short-term fix keeps using a pinned PolicyEngine/policyengine-us-data repo artifact: policyengine_us_data/storage/calibration_targets/soi_targets.csv. That avoids depending on a local PE-US-data checkout or the PE-US-data Python package, but it is still a construction-time dependency on PE-US-data-owned SOI data.

Longer term, Arch should own the source-backed IRS SOI facts and provenance, and microplex-us should adapt/export those facts into the simulator-specific PE-style SOI target surface used by PUF uprating.

Goal

Add full Arch support for the IRS SOI source data needed by Microplex PUF uprating back to 2015.

Scope

• Backfill the relevant IRS SOI source packages for years 2015 onward.
• Cover the SOI measures needed to reproduce the PE-style long target table currently consumed by microplex-us PUF uprating.
• Preserve source-level provenance, table/line metadata, units, periods, and AGI-stub semantics in Arch facts.
• Add tests that verify representative facts for each covered SOI package/year range.
• Document any source-definition breaks across years, especially line/table renames or AGI-stub changes.

Non-goals

• Arch should not choose active Microplex calibration targets or bake in PolicyEngine variable names.
• Arch should not own PE-style aging/reconciliation policy.
• The Microplex-specific export/adapter can live in microplex-us; this issue is about making the underlying historical SOI fact coverage available in Arch.

Acceptance criteria

• Arch can build source-backed SOI facts for the required Microplex PUF-uprating horizon beginning in 2015.
• The resulting facts contain enough information for microplex-us to replace its pinned PE-US-data soi_targets.csv dependency.
• Tests cover at least one current year and one historical year for each newly extended SOI package.
• Any missing or non-comparable historical SOI lines are explicitly documented with source notes.

Start by reviewing policyengine_us_data/storage/calibration_targets/soi_targets.csv and the Microplex PUF-uprating dependency described in PolicyEngine/microplex-us#155. Map the IRS SOI packages and year ranges needed from 2015 onward, then define representative current and historical fact tests. Done means Arch exposes source-backed facts with provenance, metadata, units, periods, AGI-stub semantics, and documented non-comparable lines.