Idea: Synthetic data test against non-spherical sources

[leouieda]

The tests we have so far only use spherical sources for the grains. While this is a good approximation, I imagine reviewers will ask how the method behaves for non-spherical sources. For example, the papers from de Groot's group tend to use non-spherical geometries.

There are 2 possible ways to do this:

1. The simplest: Use prisms with various geometries. This is still not fully implemented in Harmonica and would need a bit of work there to include prism magnetic fields. Limitation is that getting prisms in non-xy orientations is difficult. Also doesn't account for self-demagnetising effects due to anisotropy.
2. The most complete: Use ellipsoids with random dimensions and self-demag taken into account. The method and underlying code is at https://github.com/pinga-lab/magnetic-ellipsoid Code needs some work and possible inclusion in Harmonica as well. This is the ideal test case but the method is complicated.

Not sure what to do or it's worth doing for this first paper.