Document or decompose aggregate net_worth input

[MaxGhenis]

PolicyBench diagnostics exposed a source-data ambiguity around net_worth.

In PolicyEngine US, net_worth is a household leaf input from the data. In current public scenario exports it can be much larger than the visible asset/liability components. For example, a household may show small bank accounts, vehicle value, and auto debt, while also having a large net_worth value. That value is presumably carrying unshown home equity, retirement wealth, business equity, other real estate, or other imputed balance-sheet components.

That is defensible as a source-data aggregate, but it is easy to misuse in downstream public examples and benchmarks: models/users may treat total net worth as liquid/countable assets for SNAP or SSI even though PolicyEngine uses more specific resource variables such as bank accounts, stocks, and bonds.

Requested follow-up:

• Document net_worth as an externally supplied aggregate that is not reconstructible from the currently exposed component fields.
• If feasible, add or expose a fuller imputed balance-sheet decomposition: home value/equity, mortgage debt, retirement assets, business equity/debt, stocks, bonds, bank accounts, vehicle value/debt, other real estate, and other debts.
• Consider a test or diagnostic that checks whether aggregate net_worth can be reconciled to available balance-sheet components when those components are intended to be complete.

PolicyBench will omit net_worth from prompts for now so it does not imply policy-relevant countable assets.