Populate passive_partnership_s_corp_income (NIIT passive pass-through leg, pe-us#9306): SCF non-active-business shape, SOI/8960 level

[MaxGhenis]

What

PolicyEngine/policyengine-us#9306 adds passive_partnership_s_corp_income — the §469 passive subset of partnership_s_corp_income — and puts it in the NIIT base (§1411(c)(1)(A)(ii)). It defaults to zero, so until Microcosm populates it the engine under-taxes passive pass-through income under NIIT exactly as before. Microcosm needs a stage that assigns each record's passive share.

Design (follows the #530 doctrine: measured / derived-from-law / anchored prior)

• Not derivable from law alone: passivity is a facts-and-circumstances material-participation test (§469(h), the 500-hour and related tests); the PUF and CPS observe neither hours nor participation. So this is an anchored-prior stage, not a derivation.
• Survey shape: the SCF distinguishes businesses the household actively manages (X3103/X3104, the basis of qbi_employer_structure_v1) from holdings with no active management role (X3401-X3452) — the cleanest public survey signal for "owner is passive." Extend the v3 SCF extract to that section: P(passive holding | income band, legal form) and the passive share of pass-through income by band.
• Administrative level: SOI partnership passive/nonpassive tables and Form 8960 line-4 aggregates (chronicle ingestion companion issue) pin the aggregate passive share the assignment must hit — same SCF-shape / SOI-level rule v3 used for employer presence.
• Placement: a per-record seeded draw of passive share (own rng family, so v1–v3 streams stay byte-identical), conditioned on income band and latent entity form (the v3 partnership/S-corp split already exists), emitting passive_partnership_s_corp_income = share × partnership_s_corp_income where positive. Gate behind the version flag like every other QBI family.
• Validation: NIIT aggregate vs the JCT/SOI NIIT totals becomes a new row on the Standing per-component repeal-revenue validation surface (#298) #535 standing surface; the passive-income aggregate vs Form 8960 line 4 is the direct anchor check.

Dependencies

• Upstream: chronicle#179 (the anchors). Engine side: pe-us#9306 (open, CI green).
• Builds on the US: pin raw PUF; move all processed-PUF derivations (aging + Section 199A simulation + carried leaves) into populace #530 stack (qbi-v3-wiring); v4 of the QBI family or a sibling stage — decide at implementation.

Consistency note

Today the engine's NIIT base includes rental_income in full while Schedule E rental is often passive-by-default; the new stage should not create an asymmetry where pass-through passivity is modeled but rental passivity is asserted. Document the treatment either way.