Add documentation for configuring Julia on Casper/Derecho #364
Description
Mainly opening this as a discussion item to see if there's interest in this before I actually go writing things down.
I use Breeze.jl (which is based on Oceananigans.jl) on NCAR HPC resources. It was mentioned after recent updates that most Julia users tended to prefer managing their own software stacks and so maintaining the julia module had kind of fallen down the priority list. I was directed towards JUHPC to set up my own environment that was also aware of the HPC environment in terms of system-optimized packages for MPI and etc.
Long story short, this actually worked out quite well. I've now got separate installs for Casper and Derecho, both of which are aware of site packages for MPI and CUDA (I also got HDF5 working, but this ended up causing some other issues with Julia's NCDatasets that I think is an upstream issue, so I decided just to stick with the Julia versions of those packages since I didn't need parallel functionality). Having separate installs for each machine has helped a lot with cases where running some tasks on different machines could result in strange behavior.
However, there was definitely a few tripping hazards along the way, and it was mentioned that a lot of these steps haven't really been written down anywhere. Since this actually ended up being helpful for me, I was thinking it might be useful to write a short page that documents how I got things set up so others who go down this path have something to reference.
Does this seem like something that'd be a good contribution? If so, is there any particular bits of info that would be viewed as important to include aside from the general processes I followed?