GitLab

This change is intended to improve the behavior when multiple
CUDA contexts are created on a node with multiple CUDA GPUs,
each of which is set to process-exclusive mode.

Converted the routine for identifying molecules from a recursion to a loop.  This avoids stack overflows with very long linear molecules.

AmoebaMultipoleForce did not report particle groups, which could potentially lead to incorrect reordering

Integrators control when atom reordering gets done, instead of having it happen automatically when forces are evaluated.  This simplifies some code, and also improves unintuitive behavior with CustomIntegrator.