UK: structural reforms gated on `*.in_effect=True` silently no-op via `simulation_modifier` path

[MaxGhenis]

Flagged during subagent review of #300.

The UK model applies reforms via simulation_modifier_from_parameter_values after Microsimulation construction (src/policyengine/tax_benefit_models/uk/model.py:190,198), using p.update() on parameters. This means any UK structural reform that is gated on a *.in_effect=True parameter — where the structural reform class only activates inside Microsimulation.__init__'s call to create_structural_reforms_from_parameters — would silently no-op.

US recently hit this class of bug (#300) when gov.contrib.ctc.*.in_effect=True reforms crashed because the structural reform's variables were registered on the per-sim system but populations were built from the module-level system. The UK variant is subtly different: the structural reform never gets applied at all because p.update() happens post-construction.

Evidence

• policyengine_uk/reforms/reforms.py contains at least one in_effect gate (confirmed via grep).
• UK uses simulation_modifier at src/policyengine/tax_benefit_models/uk/model.py:187-202, which runs after Microsimulation.__init__ has already decided which structural reforms to apply (based on base-parameter in_effect=False).

What to check

1. Enumerate UK gov.contrib.*.in_effect parameters. If any exist and activate structural reforms, this is a real latent bug.
2. Write a regression test along the lines of tests/test_us_microsim_structural_reforms.py::test__gov_contrib_gate_runs_cleanly for a UK gate.
3. If confirmed, the fix is to pass UK reforms at Microsimulation(reform=...) construction time (same pattern US uses) instead of via post-construction p.update().

Not blocking v4.0.1; file as a v4.0.2 / v4.1 candidate.