kokkos: cxxstd cleanup

[tpadioleau]

• remove the always in conflict variant cxxstd=11
• add variant cxxstd=23 starting from Kokkos 4

[tpadioleau]

• Should we continue adding conflicts with CUDA, like conflicts("+cuda", when="cxxstd=23") ? I think if one wants to compile with CUDA clang++, it should work ? I think this limitation only concerns nvcc ?
• Not a big fan of the fix for HPX in C++23...

[rbberger]

Instead of operating with conflicts, we could adopt what is described here: https://spack.readthedocs.io/en/latest/packaging_guide_creation.html#conditional-possible-values

[tpadioleau]

In a previous gitlab run, there was an issue with vtk-m, likely not supporting C++20 yet (but also likely not a big bug). What should we do about it ?

[cedricchevalier19]

In a previous gitlab run, there was an issue with vtk-m, likely not supporting C++20 yet (but also likely not a big bug). What should we do about it ?

Was it with kokkos 4? If so, they should add a conflict in their recipe.
I am not sure that it is maintained now as the upstream project is now viskores.

[tpadioleau]

In a previous gitlab run, there was an issue with vtk-m, likely not supporting C++20 yet (but also likely not a big bug). What should we do about it ?

Was it with kokkos 4? If so, they should add a conflict in their recipe. I am not sure that it is maintained now as the upstream project is now viskores.

It was Kokkos 5 that triggered the use of C++20.

[lucbv]

We can reach out to the vtk-m/viskores maintainers if necessary, considering the project went through some changes they might want to clean things up and deprecate old things as well.
I like the current changes, it looks clean, thanks @tpadioleau !

[tpadioleau]

Given #2945, I am in favor to wait.

[rbberger]

@tpadioleau more cleanup: #2994

[rbberger]

@tpadioleau fingers crossed for the next rebase ;-)