Update all CI checks to also run for 2026.06

[casparvl]

I'm expecting some of these to not work (yet). I'm guessing we need some very basic stuff to be in place first. There may even be some CUDA-related GPU things that will not work yet - but let's just see how far this gets.

[boegel]

@bedroge still failing tests?

[ocaisa]

One failure is our Lmod having problems with set -e (fixed upstream), the other is because there is no cc70 in 2026.06

[bedroge]

The test checks if cc77 falls back to cc70, that doesn't work for 2026.06. That either requires:

• a change in the EESSI module: falling back to x0 is now hardcoded at https://github.com/EESSI/software-layer-scripts/blob/main/init/modules/EESSI/2023.06.lua#L213. For 2026.06 it would need a more explicit check and let any 75 < cc <80 fall back to 75.
• or shall I just override that check for 2026.06, since there is no 75 < cc <80 in practice anyway

The latter seems much easier to me.

edit: changed it to cc97 instead, which should work fine for all EESSI versions.

[bedroge]

@ocaisa is there any workaround for the set -e issue, or do we really have to update Lmod (that will take some time, would be nice to have the CI enabled for 2026.06)?