Fix a large bunch of warnings, and there are still mistakes that should be fixed before the SC release

At the user levl, we wait for a sequence. We need to add a MORSE_Barrier function that waits for all sequences currently running

some cublas kernels can be used without magma only with CHAMELEON_USE_CUDA define - fix this mistake in cudablas files

Fix build with starpu1.1 which doesn't have starpu_mpi_comm_{rank,size}: let's just call RUNTIME_comm_{rank,size} which already do things properly

add cudablas library to make calls to cuda kernels (cublas here), no more calls to cublas in runtime/starpu codelets

instead of introducing RUNTIME_distributed_barrier

add cudablas library to make calls to cuda kernels (magma here, cublas will follow), no more calls to magma in runtime/starpu codelets

Also make StarPU-MPI initialization assume that in the simgrid-mpi case we do not have any outer mpi initialization

MORSE_Distributed_size, MORSE_Distributed_rank so that applications do not
hardcode the use of MPI.

Introduce RUNTIME_distributed_rank, RUNTIME_distributed_size,
RUNTIME_distributed_barrier, so that MORSE does not hardcode the use of MPI
either.

This allows to use simgrid-mpi.