Includes a missing header fix in hxcomm@ebrains-2.0-rc3

This reverts merge request !103

This reverts merge request !102

We can drop oppulance from `spack.yaml` as it is no top-level
dependency anymore.

Importing torch yields an ImportError caused by some mismatch between
build options (distributed is enabled, but tensorpipe is not →
this seems unsupported → enable tensorpipe).

Related to: https://support.humanbrainproject.eu/#ticket/zoom/14319

Add py-quantities, which is a dependency for py-neo. This same version is available in upstream Spack, adding it here in case we need to add new versions quickly

(due to a known build fail of 1.10.x → to be investigated later)

Previously, we had compiler and python mentioned as specs and relied
on "together" concretization to pin the dependencies. Now, we provide
these constraints via "packages" section, providing global constraints.
Site-specific things, like "mpi" provider/external libs, or architecture
"target" constraints would also go in there…

See MR !75 from Eric Muller