Remove remaining calibration xfails

[MaxGhenis]

Summary

Closes #305.

This removes the remaining non-land calibration xfail markers and makes pytest xfail strict so temporary expected failures cannot silently linger.

Changes

• remove the xfail markers from:
  • policyengine_uk_data/tests/test_scotland_babies.py
  • policyengine_uk_data/tests/test_scotland_uc_babies.py
  • policyengine_uk_data/tests/test_uc_by_children.py
• enable xfail_strict = true in pytest configuration
• add a changelog fragment

Why

These tests already skip cleanly when the Enhanced FRS dataset is unavailable through the shared baseline fixture. Keeping them as xfail means CI gets no signal when the dataset is present, and non-strict xfail allows temporary expected failures to become stale.

Test plan

• uv run pytest policyengine_uk_data/tests/test_scotland_babies.py policyengine_uk_data/tests/test_scotland_uc_babies.py policyengine_uk_data/tests/test_uc_by_children.py -q
• uv run pytest policyengine_uk_data/tests/test_scotland_babies.py policyengine_uk_data/tests/test_scotland_uc_babies.py policyengine_uk_data/tests/test_uc_by_children.py policyengine_uk_data/tests/test_land_value_targets.py -q

Both runs skipped locally because the Enhanced FRS dataset is not present in this environment. CI should provide the real calibrated-dataset signal.

[vahid-ahmadi]

Two issues:

1. Lint failure: ruff formatting

All three modified test files need reformatting (likely trailing blank lines after removing the xfail + import pytest blocks). Fix:

ruff format policyengine_uk_data/tests/test_scotland_babies.py policyengine_uk_data/tests/test_scotland_uc_babies.py policyengine_uk_data/tests/test_uc_by_children.py

2. xfail_strict = true interaction with PR #304

test_land_value_targets.py::test_land_value_aggregate still has @pytest.mark.xfail (kept intentionally in #304 because it depends on recalibrated data). With xfail_strict = true, if that test unexpectedly passes (e.g. after a recalibration run), CI will fail with XPASS. That's arguably the desired behaviour (forces someone to remove the marker), but worth being aware of — especially since #304 is also open and may merge first.

Otherwise the change looks correct. The three tests all use the baseline fixture which pytest.skips when the dataset isn't available, so removing xfail is safe.