Remove property_wealth from ONS land calibration targets

[MaxGhenis]

Summary

Merged PR #292 added property_wealth as a fourth ONS land calibration target, using the same aggregate value as total land. That target is conceptually incorrect and should be removed.

Problem

property_wealth is not an ONS land-value target. In policyengine-uk, it represents total property wealth, while the ONS National Balance Sheet figure used in #292 is a land-value aggregate. Routing ons/property_wealth through the land-target computation conflates two different concepts and weakens the calibration loss.

Specifically, #292 currently:

• defines ons/property_wealth in policyengine_uk_data/targets/sources/ons_land_values.py
• routes ons/property_wealth through compute_land_value() in policyengine_uk_data/targets/build_loss_matrix.py
• treats property_wealth as if it should match the ONS land total in policyengine_uk_data/tests/test_land_value_targets.py

Expected fix

• remove ons/property_wealth from the ONS land target source
• stop dispatching ons/property_wealth through the land target computation
• keep the real land targets: household_land_value, corporate_land_value, and land_value
• keep household_land_value <= property_wealth as an invariant test, but not as an ONS calibration target

Notes

This is separate from #295. Even if land targets still fail to converge after recalibration, property_wealth should not be included as a land target.