Testing for GPU/multi-node

We need valid tests on GPU for performance and accuracy. Current implementation uses KernelAbstractions.jl. I also hear about [JACC.jl](https://juliagpu.github.io/JACC.jl/stable/) recently.

Also, we need multi-node tests for an extended scaling run.

Start by reviewing the current Julia implementation using KernelAbstractions.jl and the referenced JACC.jl documentation. Define valid GPU performance and accuracy tests, then scope the multi-node extended scaling run; the work is done when both test areas produce repeatable results.

Written by the indexing model from the issue text.