Multi-Supercomputer-Proof Generic Workflows

[CodeGat]

See https://github.com/ACCESS-NRI/build-cd/pull/111/files/a34bb0b53657f169450c8ac31a9c3463c9837567#r1758047493 and https://github.com/ACCESS-NRI/build-cd/pull/111/files/a34bb0b53657f169450c8ac31a9c3463c9837567#r1758181572

We have gone quite far in making this repo able to handle multiple supercomputer targets, but we aren't 100% there yet. Things like comment explicitly referring to Gadi, explicit paths to /g/data/vk83/... in comments, among other things.

Audit this repositories use matrices, comments and targets to make sure we can deploy to another supercomputer target as easily as possible.

One main issue is the GitHub Release - matrixing could lead to multiple releases of the same version of a model, with different spack.locks for different targets.