CI: Add PDAF builds

[jjokella]

Started from issue #81

Opening PR for testing CI.

• eCLM-PDAF: CI: Enable eCLM-PDAF build #83
• eCLM-ParFlow-PDAF: SLURM/HYPRE error
• CLM35-ParFlow-PDAF: SLURM/HYPRE error

Related PDAF-PR:

• eCLM-ParFlow related fixes in PDAF interface pdaf#30

[kvrigor]

The issue seems to be the way ParFlow linker flags are manually specified:

TSMP2/cmake/BuildPDAFMODEL.cmake

Lines 98 to 104 in f45128a

	if(DEFINED PARFLOW_SRC)
	list(APPEND PDAF_LIBS "-L${CMAKE_INSTALL_PREFIX}/lib -lpfsimulator -lamps -lpfkinsol -lgfortran -lcjson")
	# GPU
	# list(APPEND PDAF_LIBS "-L${CMAKE_INSTALL_PREFIX}/rmm/lib -lstdc++ -lcudart -lrmm -lnvToolsExt")
	list(APPEND PDAF_LIBS "-L${HYPRE_ROOT}/lib -lHYPRE")
	list(APPEND PDAF_LIBS "-L/lib64 -lslurm")
	endif()

The correct linker flags for ParFlow 3rd party dependencies (e.g. CUDA, Hypre) are generated by the ParFlow CMake scripts. Unfortunately TSMP2 cannot directly access these flags since they are dynamically generated. We need to find a way to extract these linker info from ParFlow.

[jjokella]

@kvrigor With an update in loading SLURM/HYPRE, the eCLM/CLM3.5-ParFlow-PDAF builds now complete!

Side-Note: Temporary CI failure

I get a strange ICON standalone failure, but this should not be part of this CI.

Note: ICON standalone was failing due to temprorary connection failure to https://gitlab.dkrz.de/icon/icon-model re-running these checks, they complete (as they should with no changes wrt ICON).

[jjokella]

• ParFlow (or PDAF from https://github.com/HPSCTerrSys/parflow) already contained all necessary changes
• PDAF now include
  • eCLM-ParFlow related fixes in PDAF interface pdaf#30

This turns the PR into

1. An update of the main PDAF
2. updates in the compilation scripts
   The from my side could be merged!

I am looking forward to the additional PDAF-related CI-tests!

[kvrigor]

Also Johannes, please make sure to remove Claude as a co-author before you merge the PR. It's already transparent in the branch commits that LLM-assisted tooling was used, and whatever the tools to write the code, you are the one who did the difficult work of writing up this PR. Commit authorship imply code ownership; to me a "human author" is a good signal that signifies that some human author is ultimately responsible for the changes in the code, regardless of the tools used.

I think the list of commit authors in a repo, whatever tools used, should remain human. Besides, the prevalence of "Claude/Copilot" in Git repos nowadays look like free advertising to me 😅

[kvrigor]

I get a strange ICON standalone failure, but this should not be part of this CI.

Note: ICON standalone was failing due to temprorary connection failure to https://gitlab.dkrz.de/icon/icon-model re-running these checks, they complete (as they should with no changes wrt ICON).

Let's ignore it for now.. I'll fix this flaky test after this PR is merged.