UK reform-validation suite: CenTax/HMRC/OBR published costings + the #462 backtest families as shipped payload

The UK validation surface is thin next to the US (policyengine.py#462's UK run had ~17 candidate metrics vs the US's state-legislative suite #319, JCT/OBBBA rows, and SPM backtests #348). uk-v2's #133 shows the pattern that fills it: its CenTax replication cross-checked an engine-computed reform cost against a published analysis to the pound (£28.55bn) on a defined subset.

Proposal: a UK reform-validation suite in the reform-validation payload, seeded from published, citable costings —

• CenTax published analyses (with the same attribution discipline #133 used),
• HMRC policy costings / OBR Economic and Fiscal Outlook policy tables,
• the #462 out-of-sample families (HBAI poverty, DWP caseloads/expenditure, HMRC liabilities) as baseline backtests — promoting the adjudication pack's benchmark set into the shipped payload the way #348 did for US poverty.

Out-of-sample discipline per #348/#302: diagnostics and promotion-gate evidence, never calibration targets.

Refs: policyengine.py#462, #348, #319 (US pattern), #302, policyengine-uk-v2#133.

Start by reviewing the reform-validation payload and the patterns referenced in policyengine.py#462, #348, and #302, then compare the UK proposal with policyengine-uk-v2#133. Done means a shipped UK validation suite with citable CenTax, HMRC, and OBR costings plus the specified out-of-sample backtest families, kept as diagnostics and promotion-gate evidence rather than calibration targets.