Add applications to aws-pcluster-* stacks

[stephenmsachs]

This PR adds building gromacs, openfoam, wrf, mpas-model, quantum-espresso, palace & lammps with intel, gcc and oneapi compilers in the aws-pcluster-* build cache stacks.

(This is a rebase of https://github.com/spack/spack/pull/43536 onto develop after my push/pull hickup on that PR.)

[stephenmsachs]

close and re-open to start gitlab pipeline.

[haampie]

I'm have already said that I'm not a fan, because it's ridiculously slow not to have the compiler toolchain cached in a docker image:

9 minutes 34 seconds 

is the time it takes to "build" ca-certificates-mozilla, which has an install time of less than a second.

I'm sure you can find a way to put some compilers in a docker image in a way that's technically not redistribution. Either a private image somewhere, or some other way to signal that the image is not meant for redistribution, but serves as a cache.

Also note we already use ghcr.io/spack/ubuntu22.04-runner-amd64-oneapi-2024.0.0:2024.01.16b which contains intel compilers.

[stephenmsachs]

I am looking for alternatives to speed up the container bootstrap time.

Also note we already use ghcr.io/spack/ubuntu22.04-runner-amd64-oneapi-2024.0.0:2024.01.16b which contains intel compilers.

That container is built from an Intel provided container. Currently, there is no Intel provided container for Amazon Linux2.