(closes #2629) Fix NEMOv5 issues

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 99.86%. Comparing base (92fbf60) to head (07fec74). Report is 18 commits behind head on master.

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #2698 +/- ## ======================================= Coverage 99.86% 99.86% ======================================= Files 353 353 Lines 48996 49024 +28 ======================================= + Hits 48932 48960 +28 Misses 64 64 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[sergisiso]

@arporter This is ready for review, it compiles NEMOv5 and LFRic with the new spack repository, where there is software stack to build the applications with gfortran-14 nvhpc-24.5 and intel-24.2.

I addition:

• I updated the version of LFRic, which we now compare using the produced checksum file, so it can be tested with MPI. (note that before we compiled with mpi but run serially which could hide some bugs)
• I deleted NEMOv5 build system logic to exclude files in order to have them all in the provided script. This allows us to be explicit about what is failing, and fix the problems without having to update the files inside gh_runner manually.
• I disabled the addition of parallelisation directives inside pure or elemental functions, this produced a compiler bug in NEMOv5 and it seems to be invalid by definition.
• I fix an issue where we lost the pure/elemental function attributes

I have not yet deleted the current NEMOv5 version that there is in the nemo_test workflow, because the output is very slightly different: it : 10 |ssh|_max: 0.6483453724882898D+01 |U|_max: 0.5701938594625860D-02 |V|_max: 0.4920016849737715D-01 S_min: 0.2996960233984670D+02 S_max: 0.3102375152368392D+02 it : 10 |ssh|_max: 0.6483453724938688D+01 |U|_max: 0.5701938594606864D-02 |V|_max: 0.4919999106597320D-01 S_min: 0.2996960234052442D+02 S_max: 0.3102375152367826D+02

If the reviewer I ok with this numbers I could update the KGO file, do the nemov5 check for this compiler with exact output match and deleted from the nemov4 workflow.

[arporter]

If the reviewer I ok with this numbers I could update the KGO file, do the nemov5 check for this compiler with exact output match and deleted from the nemov4 workflow.

If this change in the pass-through results is purely because of a change in the compiler version then I'm happy. Probably worth checking that the model is stable for a longer run though (perhaps 500 steps).

[sergisiso]

@arporter We are now down to 1 NEMOv4 passthrough error (#717) and 0 NEMOv5 passthrough errors! Although I am wondering if we should enable TOP and ICE at least in the passthrough in order to cover more code.

[sergisiso]

@arporter This is ready for next review

[sergisiso]

I just realised that I made a typo in the branch name and some commits by referring to 2692, but this PR solves issue 2629, sorry for the confusion

[sergisiso]

@arporter Ready for another review, I just slightly updated the test filenames and removed del_keys on the workflow.