Add structural reform support for AI agent

[nikhilwoodruff]

The AI agent needs to write and run structural reforms (using PolicyEngine's simulation modifier attribute) on UK and US simulations.

What's needed

Modal functions:

• simulate_household_uk_with_modifier / simulate_household_us_with_modifier - household sims with custom variable logic
• simulate_economy_uk_with_modifier / simulate_economy_us_with_modifier - economy sims with modifiers
• economy_comparison_uk_with_modifier / economy_comparison_us_with_modifier - full comparisons with structural reforms

Agent-facing upload endpoints:

• POST /policies - already exists, may need modifier field
• POST /simulations/results - agent uploads completed simulation results
• POST /households/results - agent uploads household calculation results
• POST /analysis/decile-impacts - agent uploads decile impact data
• POST /analysis/poverty - agent uploads poverty metrics
• POST /analysis/inequality - agent uploads inequality metrics

Data model changes:

• Add simulation_modifier field to Policy or create new StructuralReform model
• Store Python code that defines custom variable formulas

Context

PolicyEngine's simulation modifier allows defining custom variable formulas at runtime - essential for structural reforms like changing eligibility rules or adding new benefits. Currently the API only supports parameter value changes. The agent needs to both run these simulations and write back results directly.