Add cash ISA and stocks and shares ISA holdings variables

[vahid-ahmadi]

What

Adds two household-level wealth input variables:

• cash_isa — amount held in cash ISAs
• stocks_and_shares_isa — amount held in stocks & shares (investment) ISAs

Both are Household, STOCK, GBP, GDP-per-capita uprated, mirroring corporate_wealth / savings.

Why

The only ISA variable today is the income flow individual_savings_account_interest_income. ISA balances had nowhere to land:

• investment ISAs were folded into corporate_wealth in the (now archived) policyengine-uk-data wealth imputation, so they were not separable; and
• cash ISAs were not represented at all.

These variables provide the model-side home for ISA balances imputed by the data pipeline, enabling ISA-balance modelling (ISA reforms, wealth-tax scenarios).

Data side

The matching data work — surfacing cash_isa and stocks_and_shares_isa from the Wealth and Assets Survey in the populace UK build — is tracked in PolicyEngine/populace#180.

[MaxGhenis]

Approving (agent-assisted review). The variables are correct and cleanly scoped: cash_isa and stocks_and_shares_isa mirror corporate_wealth/savings exactly (Household, YEAR, float, GBP, OBR per-capita GDP uprating, STOCK), and they're pure inputs with no consumers anywhere in policyengine_uk/, so nothing existing moves. No double-count either: total_wealth still adds only property/corporate/savings — since the data side folds S&S ISA into corporate_wealth, exposing it separately is fine, and cash ISA staying out of total_wealth matches its prior (unrepresented) treatment.

One mechanical fix I pushed to the branch rather than bouncing back: the root changelog_entry.yaml is inert since the towncrier migration (#1506) — bump_version.py and the versioning workflow only read changelog.d/**, so merged as-is this would have gotten no version bump and no changelog line. I swapped it for changelog.d/add-isa-balance-variables.added.md (the .added suffix is what makes the bump come out minor, matching your intent).

Sequencing note for the populace side: populace#181 doesn't depend on this at build time (spec-only), but the eventual UK runtime PR that emits these columns should bump the uk extra to the release carrying these variables.