Disable test(s) that fail often on GitHub CI for Windows/MPI

[sloede]

In the past weeks/months, we have a very high failure rate for CI tests on Windows with MPI that are intermittent and cannot be traced back reasonably to any real programming issue. Some examples:

https://github.com/trixi-framework/Trixi.jl/actions/runs/4703123828/jobs/8341232910
https://github.com/trixi-framework/Trixi.jl/actions/runs/4710627754/jobs/8362973268
https://github.com/trixi-framework/Trixi.jl/actions/runs/4714298641/jobs/8360570746
https://github.com/trixi-framework/Trixi.jl/actions/runs/4698811234/jobs/8351234657
https://github.com/trixi-framework/Trixi.jl/actions/runs/4694820492/jobs/8323339119

Our working hypothesis is that the GitHub runners run out of memory, which is supported by at least some of the failed tests, where we see the following statement in the test failure overview:

This happened, e.g., here:
https://github.com/trixi-framework/Trixi.jl/actions/runs/4702374974

The goal of this PR is to try to disable those tests that only fail on Windows to have regularly passing MPI tests again, and then fix this for real (e.g. by reducing the test size) in a subsequent PR.

[codecov]

Codecov Report

Merging #1398 (8c68ddf) into main (d952106) will increase coverage by 0.54%.
The diff coverage is n/a.

❗ Current head 8c68ddf differs from pull request most recent head 08f3a96. Consider uploading reports for the commit 08f3a96 to get more accurate results

@@            Coverage Diff             @@
##             main    #1398      +/-   ##
==========================================
+ Coverage   95.44%   95.97%   +0.54%     
==========================================
  Files         351      351              
  Lines       29122    29122              
==========================================
+ Hits        27793    27949     +156     
+ Misses       1329     1173     -156     

Flag	Coverage Δ
unittests	95.97% <ø> (+0.54%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

see 6 files with indirect coverage changes

[sloede]

Alright... after a lot of trial and error I am now relatively confident that this represents a working set of MPI-parallel tests for GitHub Actions on Windows. The good news is that this set has been run successfully for at least 4 times in a row without fail. The bad news? I had to completely disable all parallel tests of the P4estMesh.

Where to go from now? My current assessment is still that there is might be a fundamental problem with memory usage However, this is something which I believe less and less. Alternateively, there might be an issue with p4est on Windows, or there is a more fundamental issue with MicrosoftMPI. Neither cause can fully explain why the failures were intermittent, only affected some tests and not all of them, and why also TreeMesh tests are failing.

My suggestion for how to proceed would be to merge this PR (possibly after cleaning up the p4est tests file, since it can be excluded in toto), and then to create an issue such that this can be investigated more thoroughly in the future. Thoughts, suggestions?

[ranocha]

Thanks a lot for looking into this issue! I just have a minor suggestion to discuss.

[ranocha]

Thanks!