Add opt-in cliff impact outputs to reform analyses

[anth-volk]

Summary

Add an opt-in cliff impact output to US and UK reform analyses, matching the legacy simulation API shape:

{
  "baseline": {"cliff_gap": ..., "cliff_share": ...},
  "reform": {"cliff_gap": ..., "cliff_share": ...}
}

Motivation

The simulation API supported cliff impact outputs on policyengine.py < 1. The new v4 analysis path needs a compatible way to materialize and return cliff impact calculations without making the expensive cliff variables part of every default simulation output.

Proposed behavior

• Add a public cliff impact output helper.
• Materialize cliff_gap, is_on_cliff, and is_adult only when requested.
• Expose include_cliff_impacts on US and UK economic impact analysis.
• Preserve existing labor supply response extra-variable behavior.

Validation notes

Local parity validation against the legacy simulation API on pinned package/data versions found a nonzero sampled cliff case and matched the legacy aggregate exactly:

• cliff_gap: 3,952,908,577.8710938
• cliff_share: 0.0060240964024626835

Focused tests also passed locally:

• pytest tests/test_cliff_impact.py tests/test_cliff_impact_analysis.py
• pytest tests/test_labor_supply_response.py tests/test_uk_analysis.py tests/test_uk_program_statistics.py
• ruff check on changed source/test files