TODOs for natESM sprint

[wiltonloch]

This is a non-exhaustive list of the tasks still remaining for the natESM sprint on CLEO (likely in priority order):

Completed:

• [x] Fix the distributed memory parallelization (currently there are differences in the results when the domain is uneven)
• [x] Fix the parallel output (this is likely related to the previous point, corrupted output binary files)
• [x] Add CI run tests for the parallelization
• [x] Improve the inter-process communication in move_supers_in_domain to not use global collective calls

On the YAC coupling:

• [ ] Fix bubble test case bug in wind fields with one-way YAC coupling
• [ ] allow YAC coupling with different gridbox sizes in x and y
• [ ] Add CI run tests for the coupling
• [ ] Update the PR on ICON for the CLEO coupling and make it ready for review

On the MPI Domain Decomposition:

• [ ] Add arrays for the gridbox sizes in the cartesian decomposition to allow gridboxes with different sizes (consistent along the other dimensions)
• [ ] Make send/recv of super-droplets moving between gridboxes on different processes GPU build compatible
• [ ] Improve the logging so that just one process does the printing
• [ ] Integrate the distributed memory parallelization with the YAC coupling
• [ ] Do a simple performance evaluation with a large enough experiment
• [ ] Turn the vertical penalization into a config parameter when finding the best decomposition
• [ ] Improve the factorization algorithm, so that there is no need for a special case when creating the decomposition

[yoctoyotta1024]

Also important note to remember that currently (WIP to generalise):

• grids for CLEO using YAC assumed to have equal x and y grid spacings
• grids for CLEO using MPI assumed to have equal x, y and z grid spacing
• MPI only for 3-D non-GPU builds