GuidingCenter en_fB sums over holes and ghosts; test_scalars en_fB tests fail on devel #727

Two tests in src/struphy/models/tests/test_scalars.py fail on devel (25688c5):

FAILED src/struphy/models/tests/test_scalars.py::test_en_fB_ignores_ghost_markers
  AssertionError: GuidingCenter: en_fB = 8.75 with ghosts, 2.0 without
FAILED src/struphy/models/tests/test_scalars.py::test_guiding_center_en_fB_does_not_divide_by_Np
  AttributeError: '_GuidingCenterMarkers' object has no attribute 'save_magnetic_background_energy'

The failures were first seen while testing #724, but they also happen on unchanged devel (checked locally in a clean worktree of origin/devel, macOS, python -m pytest src/struphy/models/tests/test_scalars.py: 2 failed, 4 passed). CI does not flag them.

1. GuidingCenter._compute_en_fB uses an inverted mask (a real bug). src/struphy/models/guiding_center.py:120-121:

energy = particles.markers[~particles.valid_mks, 5].dot(
    particles.markers[~particles.valid_mks, 8],
)

~valid_mks selects the holes and ghost markers instead of the valid markers, so en_fB of GuidingCenter sums over the wrong rows. #631 (fix for #563) meant to change ~particles.holes to particles.valid_mks. It did so in LinearMHDDriftkineticCC._compute_en_fB (linear_mhd_driftkinetic_cc.py:211-212), but in guiding_center.py it wrote ~particles.valid_mks. test_en_fB_ignores_ghost_markers catches this.

2. The mock in test_guiding_center_en_fB_does_not_divide_by_Np is outdated. _GuidingCenterMarkers (test_scalars.py:169, added in #477) has no save_magnetic_background_energy() and no valid_mks, which GuidingCenter._compute_en_fB uses.

Suggested fix:

• guiding_center.py: index with particles.valid_mks (without ~).
• test_scalars.py: give _GuidingCenterMarkers a no-op save_magnetic_background_energy() and a valid_mks attribute (~self.holes).
• Check why CI did not run these tests. The unit-test job uses --testmon, so it may have skipped them.

Found while working on #724 (part of #697, #650).

🤖 Generated with Claude Code