Create the SPI income channel on the microcosm spine with the FRS-only second stage #683

Workstream E7 of the UK-migration master epic #665; child of #145.

Why

The existing hmrc_spi_income stage restores the SPI channel inside the certified incumbent H5. Once the microcosm spine exists (E2–E4), the stage must create that channel itself.

Scope

• Stack the 10,000 zero-weight SPI-synthetic households (household_is_spi_synthetic; the zero-weight prior-mass-share 0.5 semantics preserved for the calibration seam) — prefer a generic stack_zero_weight_donors operation.
• Stage-1 QRF (predictors age/gender/region only — already ported) imputes the income surface on the synthetic block; the real frame gets only dividend_income re-imputed (incumbent behavior).
• The FRS-only stage-2 QRF re-imputes the 29 FRS-only person variables on the SPI donors (stops a high-income imputed donor carrying its FRS donor's benefit receipts) — already declared in the manifest; the work is the channel-creation path plus coverage-manifest family updates.
• The 5 SPI leaves with no raw-FRS source (EPB, EXPS, TAXTERM, MOTHINC, OTHERINC) remain evidence_absent — never proxied.
• Coordinate with National staging rebuild fails the current terminal gates: constant source_year release column and SPI weight tail above the reviewed maximum #630 as we go: the SPI channel is where the weight-ratio tail lives; if National staging rebuild fails the current terminal gates: constant source_year release column and SPI weight tail above the reviewed maximum #630 resolves by repairing the tail, this port inherits the repair.

Acceptance

• The 208-fact HMRC replay reproduced on the microcosm spine with the same fence structure (or better, if newly-retained raw constituents unlock bands).
• gift_aid / charitable_investment_gifts clear the effective-mass floor.
• Zero-weight strata declarations re-derived at the grain adjudicated in the spine-grain issue (pre-clone 10,000 vs post-clone 100,000).

UK-specificity note (required in the PR)

SPI band semantics are UK (and already fenced); the donor-stacking and two-stage QRF operations should be generic. List anything beyond that and why.

Depends on: E2–E4 (the full FRS-only variable surface that stage 2 draws from); the spine-grain adjudication.