Minimal petsc4py code for ksp

[enricofacca]

Minimal code using petsc4py and ksp solver for linear systems.

Test add to test_wasserstein

[jwboth]

Thanks a lot, Enrico! This looks really good. Most comments are curiosity questions, me wanting to better understand what is going on.

Most critical is the failing installation of DarSIA including petsc4py. The style (black etc.) will sort out. I can take care of that once the installation is fixed.

[enricofacca]

Some comments on why we are not able to install petsc4py.
Reading this we first need to install numpy (than has recently being updated to version 2.0) and mpi4py.
Than we should install petsc and petsc4py via pip.
I was able to make it work but mumps and hypre are missing.
It should be possible to include this component using environment variable (as described here) but it does not work.

I was able to make everything work using conda + create env + installing petsc with conda install (pip install inside the env does not work). Hypre and mumps are, by default, included in the conda installation of petsc.

[jwboth]

I removed petsc4py and cython from the requirements. Instead, now petsc4py (at least on github actions) is build in the build workflow, see #374 . The installation for Windows will not fully work from command line I fear, and not just using pip. I am trying therefore instead to set up a separate routine for that. Can you check, whether you can run the installation of petsc incl. hypre and mumps as done in the ci script?

[enricofacca]

I answered your points. But I just wanted to test the CI. I am not sure I did the review as it supposed to be.

[enricofacca]

Testing works but it took "30 minutes" to complete with the inclusion of hypre and mumps.
This may be a problem if there are limitations on free testing time on Github.

We can just use native petsc solvers (replacing "hypre" with "icc" and mumps with the standard LU factorization), include petsc. So that the building test is passed and it takes way less time.

However, the installation of mpi4py requires MPI being installed. This may cause problems to some users..

[jwboth]

Looks good for now! Thanks a lot!