doc: added tutorials with cupy+mpi

[mrava87]

This PR adds a new directory called tutorials_cupy which includes tutorials running with CuPy arrays and MPI communicators.

Following what previously done for tutorials_nccl, a couple of new targets are also added to the Makefile to allow building the documentation with MPI or MPI/Cupy+MPI or MPI/CuPy+MPI/CuPy+NCCL tutorials.

[rohanbabbar04]

Using rank % device_count seems appropriate, as it allows flexibility in the number of ranks without requiring an equal number of GPUs.

Otherwise, it looks great—let me know if you agree with me 🙂

[mrava87]

Using rank % device_count seems appropriate, as it allows flexibility in the number of ranks without requiring an equal number of GPUs.

Otherwise, it looks great—let me know if you agree with me 🙂

Thank @rohanbabbar04! Yes, I agree, that is a good point 😄 I am going to do give quick look at all the tutorials and then merge