GitLab

Refactored ForceImpl, lots of KernelImpl subclasses, and other related classes to avoid redundant calculations when requesting a State with both forces and energies

Exported some symbols from Cuda plugin, for use by external NML plugin.
  cudaOpenMMInitializeIntegration, gpuSetConstants, CudaPlatform
  (but I couldn't figure out how to export kGenerateRandoms)
Also turned off TestRandomCuda on Windows, since it always fails and makes the dashboard look bad.
Renamed InitializeIntegration method to cudaOpenMMInitializeIntegration, now that symbol is being exported.

Free energy plugin added
Plugin will not run w/ Obc or GB/VI forces unless line 2004 of gpu.cpp (gpu->sim.totalNonbondOutputBuffers  = 2*gpu->sim.nonbondOutputBuffers;) is commented in -- working on removing this constraint
Also unit tests for GB/VI currently fail 

Restructured how force and energy computations are done to make it easier to write plugins for the CUDA platform

Optimization: introduced a kernel to clear forces at the start of each time step rather than relying on the stream

 * capitalization of gpuTypes.h
 * various cmake issues

Changed how kernels get initialized.  StandardMMForceField now has you set the 1-4 interactions explicitly.  CMMotionRemover allows the removal frequency to be set.