numpy-like assignation for Grid1f and Grid3f #536
Description
Please describe what you want to do with CRPropa and you currently can't
I need to compute several density grids for source sampling over large volumes. The grids are not, in principle, periodic. I calculate the values using numpy to save time. But assign the values to a CRpropa Grid1f is fast enough but it is not optimal for very large grids
Describe the solution you'd like
I would like to have a method similar to the index slicing in numpy for value assignment in Grid1f and Grid3f
Describe alternatives you've considered
Once I got the calculation of my PDFs over a grid with numpy (taking 2s at most using just a cunk approach), I do need to assign these values to a CRpropa Grid1f instance. For a grid with 512x512x512 cells, the total runtime is 2 minutes to create the CRpropa grid. That is not bad, but it is a bottleneck if I require a smaller step to estimate the cells. I've been considering a multiprocessing with shared memory approach (as I am assuming that Grid1f is not thread-safe and I already tested that the object cannot be pickled)
Additional context