Moved ForceEnergy into useful.h, added atomicAdd for ForceEnergy

Commented unwanted conditional in ComputeKernel that made simulations using the softcore code include self-interactions

Made rigid body grid--grid kernels use atomicAdd to communicate force/torque for entire kernel instead of each block writing data

(1) Protect dynamic memory allocations with conditionals to prevent zero-byte allocations

(2) Correctly initialize device data for particles_d using sizeof(int) instead of sizeof(int*)