Port UK local target surface and credibility gates from uk-data

[MaxGhenis]

Parent: #144.

Why

Populace has a UK local target surface, but it is smaller than the late policyengine-uk-data target stack. Before archiving uk-data, local targets need an explicit Populace home and gates that can say whether constituency/local-authority estimates are credible.

Scope migrated from uk-data

Open/recent uk-data items:

• Constituency weights produce unrealistic rent estimates for Scotland and Wales policyengine-uk-data#316: Scottish/Welsh constituency rent estimates too smooth/low.
• Wales constituency results seem a little low policyengine-uk-data#206: Wales constituency results low.
• Regional household land values calibrated ∝ property wealth — need region-specific land-to-property ratios policyengine-uk-data#357: regional household land values need region-specific land-to-property ratios.
• Add National Insurance targeting policyengine-uk-data#88: National Insurance targeting.
• Reinstate dfe education spending test policyengine-uk-data#199: DfE/education spending target test if still in the export surface.
• Adjust VAT methodology to include flat % assumptions on VAT band shares by consumption category policyengine-uk-data#241: VAT methodology and band-share assumptions.
• Recent target PRs SIPP+SCF blend + SSI-aware refit: land the asset magnitudes for the Build J re-certification #374/Build L epic: ACS second support spine via warm-start reselection #390/Release contract needs dataset-role classes: national-default checks reject valid non-default local-area artifacts #398/Bound QRF leaf storage for near-discrete targets #415/Gate the SSI swap delta at a runaway cap, not assignment granularity #431/Extend export-mass parity exclusions for the Build M frame #433: council-tax band/net targets, refreshed source vintages, rail aggregates, missing-source masking, bus targets, and public-sector employment target.

Current Populace gap

populace.build.uk.local_targets currently covers HMRC employment/self-employment amount/count, ONS age bands, UC household/children rows, and selected LA income/tenure/rent rows. It does not yet cover the full late uk-data surface or sparse missing-target semantics.

Acceptance criteria

• Target definitions are named and documented against PolicyEngine-UK variables they calibrate or validate.
• Missing local target cells can be masked deliberately; zero-valued targets remain intentional rather than silently dropped.
• Local gates include salient checks for rent in Scotland/Wales, Wales constituencies, council tax, public-sector employment, bus/rail, NI, and regional land values.
• Release diagnostics identify weakest target families and weakest local areas, not just global final loss.
• Any target not ported has an explicit rationale in the Populace target registry/manifest.