PUF donor uprating overstates mean-reverting distribution series: non_sch_d at 2.39x the SOI concept

Found adjudicating Build N attempt 2 (#462 thread). Measured on the sha-pinned puf_2024.h5 donor via puf_tax_unit_donor_from_arrays:

The eCPS-era pipeline produced $13.69B / 4.01M units from the same PUF lineage (recorded in #462's motivation), so the current ETL's 2024-level uprating roughly doubles what the retired pipeline carried. Capital-gain distributions are a mean-reverting series (TY2015 — the PUF base year — was a historically heavy distribution year); uprating its level by a generic growth factor overstates the modern concept on both margins.

Interim: #486 pins the column's level to the ledger-fed aged Table 1.4 dollar fact at release time (the SS-components repair precedent), with the factor recorded in support_value_repairs. That is honest but uniform — it preserves the donor's (too-rich) shape.

Root fix: variable-specific donor uprating for distribution-class series — age E01100-class amounts by their own SOI chain (the #474 tips mechanism generalized) either in the irs-soi-puf ETL or at populace donor load, with the chain facts in the ledger. Candidates beyond non_sch_d: the other Table 1.4 rows the #464 blanket watches, and the 13 tail-concentrated sparse columns from the N attempt-2 battery (those also implicate #481's weighted-leaf-draw loss).

Relates #462, #464, #481, #486.

Start by reviewing puf_tax_unit_donor_from_arrays and the current irs-soi-puf ETL or populace donor load path, then compare the donor figures with the ledger facts described in #486 and the findings in #462. The fix is complete when distribution-class series use variable-specific SOI uprating with chain facts recorded in the ledger, including coverage for the related Table 1.4 and sparse-column cases.