MPI refactoring

• removed __cuda_callable__ from DistributedNDArray and DistributedNDArrayView methods
• removed MpiCommunicator completely (remains only in the Problems and Solvers machinery)
• removed wrapper functions AllGroup, NullGroup, NullRequest
• renamed getCommunicationGroup to getCommunicator in all data structures
• Array.h, ArrayView.h: replaced send, receive with MPI::send, MPI::recv
• String.h: replaced mpiSend, mpiReceive with MPI::send, MPI::recv (same as above)
• removed print method and operator<< for DistributedArray and DistributedArrayView (it does not make sense to print a distributed array like this, because it may contain ghost elements)
• fixed forElements method in DistributedArray and DistributedArrayView