[Bug]: FAENet uses a different periodic boundary condition implementation

[laserkelvin]

Expected behavior

Models are expected to re-use the same pipeline for computing periodic boundary conditions/images - i.e. matsciml.datasets.transforms.PeriodicBoundaryConditions and subroutines.

Actual behavior

Per this line of code in FAENet.read_batch, it actually calls the non-pymatgen version of the PBC calculation.

Steps to reproduce the problem

N/A

Specifications

N/A