halo exchange integration

[TomMelt]

Fixes #132

Change Description

This PR adds necessary changes to support halo cells into nextsimDG.

• ModelArray's are padded with halo cells at the moment they are read from disk
• This happens in the appropriate IO routines, e.g.,
  • core/src/ParaGridIO.cpp
  • core/src/RectGridIO.cpp
• Similarly when data is written to disk the halo cells are discarded

This PR does not add halo exchange logic into nextsimDG. This will be done
in a following PR. But all of the functionality is here and it is tested in
core/test/HaloExchangeCB_test.cpp and core/test/HaloExchangePB_test.cpp
(closed and periodic boundaries).

As part of this work, 2 singleton classes have been added:

• ModelMetadata (already existed but has been refactored as a singleton)
• ModelMPI (new class)

ModelMPI stores the MPI communication, rank and size. ModelMetadata contains
metadata on the Model which includes domain decomposition information which is
required for halo exchange.

Domain decomposition data is produced by the domain decomp tool, which is then
read by ModelMetadata when it is first initialized.

I have tried to privatize as much of the class data/methods as possible to
simplify their APIs and avoid mis-use. Therefore member data is privatized as
much as possible and public getters have been made available.

Test Description

This PR adds 4 new tests to check the new functionality, for closed and periodic
boundaries (CB/PB respectively)

• ModelMetadataCB_test.cpp and ModelMetadataPB_test.cpp - test
  getPartitionMetadata can correctly load metadata from a partition metadata
  file into the appropriate data structure e.g., neighbourExtents, etc.
• HaloExchangeCB_test.cpp and HaloExchangePB_test.cpp - check that we
  can do basic halo exchange by making use of the new slicing utilities e.g.,
  ModelArraySlice

[jwallwork23]

Note that I changed the logic for determining whether the Docker workflow is run. I an attempt to fix the test failure https://github.com/nextsimhub/nextsimdg/actions/runs/12009927720/job/33475689562 might it be feasible to rebase on top of or merge in develop?

Compare https://github.com/nextsimhub/nextsimdg/blob/halo-exchange/.github/workflows/docker.yml vs https://github.com/nextsimhub/nextsimdg/blob/develop/.github/workflows/docker.yml.

[jwallwork23]

Nice that function signatures are more consistent between the MPI and non-MPI cases with this change.

I've not gone through all of the new test code but I have a few comments and requests on the source code and changes to the XIOS tests.

I think some of the changes you made to ParaGridIO.cpp need to be reflected in ParaGridIO_Xios.cpp, but that's my problem, not yours. I'm working on the relevant bits of code in #916 and #917 anyway.

[jwallwork23]

Thanks for all your hard work on this, @TomMelt! It mostly looks great but I have a few minor requests - mostly related to docs and rewording. Then it should be good to go.

I also have a few suggestions for code changes to ease understanding and reduce duplication but I think they can wait until after this is merged.

[jwallwork23]

Thanks for addressing those @TomMelt. Let's merge it! 🥳