Update aws-pcluster-amazonlinux-2 stacks

[lrbison]

Note

This PR was migrated from spack/spack#50719.

These stacks start the build with gcc@12 installed from a local buildcache

DNM: this is still in testing, and uses a runner image from a PR rather than a tagged verison.

Depends on: spack/gitlab-runners#66

[lrbison]

The build timed out after 1-hour, although it seems to have only had about 90 seconds to execute the container. I suspect significant time was spent in this mode:

Waiting for pod pipeline/runner-glrtr-zc-project-57-concurrent-6-qc8a4oz1 to be running, status is Pending
	ContainersNotReady: "containers with unready status: [build helper]"
	ContainersNotReady: "containers with unready status: [build helper]"

--
I made a no-op push just to make the CI run again.

[lrbison]

This isn't updating the build like I think it should. Nothing is being rebuilt. I will close and re-open another one.