gh-actions: update to intel oneapi 2025.3

[marcfehling]

This patch updates the github runners to the Intel 2025.3 stack.

The linux-gnu-openmpi job does not seem to compile with [email protected]. In #14811 (comment) it didn't work with [email protected] either. Please investigate @cxp484.

[marcfehling]

I'm getting an error in the linux-intel-intelmpi job with the latest compiler. It looks like it doesn't like the parameter for specifying the fortran standard.

mpiifx -c -check all -warn all -diag-error=remark,warn,error -O0 -g -traceback -fpe0 -nofltconsistency -stand:f23 -no-wrap-margin -fpp -DGITHASH_PP=\"-oneapi-2025.3\" -DGITDATE_PP=\""Wed Dec 17 15:43:31 2025 +0100\"" -DBUILDDATE_PP=\""Dec 17, 2025  14:46:18\"" -DWITH_MKL -I/opt/intel/oneapi/mkl/2025.3/include   -DUSE_IFPORT  ../../Source/imkl.f90
ifx: command line error: ignoring option '-stand'; argument must be separate

Any idea what this is about? See also:
https://www.intel.com/content/www/us/en/docs/fortran-compiler/developer-guide-reference/2025-3/stand.html

[drjfloyd]

In the most recent compiler the format for the stand option changes. It is now a space rather than a colon.

[drjfloyd]

I think NIST still has the prior version which needs the colon. Kind of an annoying change to have made between releases.

[marcfehling]

In the most recent compiler the format for the stand option changes. It is now a space rather than a colon.

Thanks! I have adjusted the parameter accordingly. Let's see if the testers are happy.

According to the official documentation, the colon option was never supported for neither linux nor macos, but only for Windows? I am surprised that the compiler did not complain earlier...

Before merging this PR, please make sure that the command line options are working on your software stack at NIST!

[cxp484]

It is working at NIST cluster. So merging the changes.