[WIP] Add an HPCToolkit stack

[blue42u]

@tgamblin @jmellorcrummey @wyphan @mwkrentel
This is the near-finalized version of #31598.

We'd like to provide a publicly available build of HPCToolkit others (and ourselves) can easily install and use. We'd also like to make sure HPCToolkit still compiles as our dependencies continue to update so we can stay on the bleeding edge. Adding a "stack" to the set of environments maintained by Spack seems to be the right choice to provide these.

This stack contains builds of HPCToolkit for 5 publicly-available operating systems (CentOS 7, Alma Linux 8, Ubuntu 20.04, OpenSUSE Leap 15, Fedora 36) and 3 generic architectures (x86_64_v3, aarch64 and ppc64le). For each enabled combination of this matrix, this stack contains builds supporting NVidia and AMD GPUs, and our WIP support for Intel GPUs, as well as a CPU-only build with less dependencies. All other settings are more-or-less copied from the E4S stack.

I've marked this as "[WIP]" as there are a few pending changes that could (but do not need to be) addressed before merge:

• The Power (ppc64le) builds are currently disabled as the Spack CI fleet does not currently include Power machines.
• The images used for most OSs are personal builds instead of "official" images from the gitlab-runners repo. The Dockerfiles used are included in Add new images for various operating systems gitlab-runners#6.
• The gitlab-ci:mappings is obscenely large due to the large matrix of supported configurations. Support full combinatorics in gitlab-ci:mappings #32290 will help with this.

[naromero77]

Does this PR need more work?

[blue42u]

It does, the main remaining workitems are to:

• Update for the drastic CI stack changes from a few months ago (CI boilerplate reduction #34272),
• Set the new python variant for hpctoolkit (either +python or ~python), and
• Reconsider the set of supported OSs (RHEL 9 and Fedora 38 are out now, CentOS 7 is pretty much defunct, Ubuntu 22.04 might be good to include, etc.).

None of those are difficult tasks, but I (continue to) have higher priority work on my list. There also hasn't been much internal push on my end to get this finished.