Create particle data structures on the device and data transfer tasks to map the transformations.
We need to move from AoS on the host to SoA on the device. Will be programmed as the only device existing (multiple GPUs inside the same MPI process will not be considered at this time).