UK CGT stage: HMRC outturn level anchors + Advani-Summers shape, with accrued/realized/taxable name separation #364

uk-v2's cgt_impute.py anchors imputed capital gains to HMRC CGT outturn statistics; populace's declared roadmap for the UK CGT stage is the Advani-Summers distribution. These are complements, not rivals: A-S gives the within-distribution shape, HMRC outturn gives administratively-observed level anchors — exactly the two-part pattern Build H just proved for the US (SOI Table 1.4/2.1 level targets identifying imputed dimensions, #299).

Proposal for the UK stage (#145 scope):

• Adopt HMRC CGT outturn (total gains, taxpayer counts, by-band where published) as ledgered calibration/referee targets — the UK mirror of the US SOI identification.
• Use A-S for the imputation model's distributional shape; the review of policyengine-uk-v2#133 verified uk-v2's income-centile-ranking approach works and is a reasonable v1 implementation reference.
• Reserve distinct variable names for accrued vs realized vs taxable gains up front (the Capital gains as primitives: asset-class positions + accrued gains + realization model, with dual-mode provenance #320 primitives taxonomy) — uk-v2's review flagged a capital_gains concept-collision risk; populace should never have one artifact where the same name means different concepts across stages.

Refs: #145, #320, #299 (the US precedent), policyengine-uk-v2#133.