Create separate communicators for each system

Description

This is the first step towards parallelization over systems. Each system takes a mpi_grp that is then internally passed to mc and further split for the different parallelization strategies.

At the moment, the communicators are simply duplicated. To achieve parallelization over systems, the communicators need to be split while iterating over the system list of a multisystem.

Fixes #345 (closed).

Checklist

• I have checked that my code follows the Octopus coding standards