Finite Build Coils

[ryanwu4]

Resolves #1169 with implementation specifically for rectangular cross section self field. Includes class implementation for FourierPlanarFiniteBuild. Test implementation is still work in progress, inclusion in coil objective function is also work in progress. Includes a garbage jupyter notebook just to show how the function works.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[dpanici]

See CoilCoilMinDistance objective for masking logic for avoiding doing biot savart of a coil over itself.

[ryanwu4]

I identified the 2 issues preventing JAX grad:

1. safenorm is necessary if eval points and coil points are the same, which is the case for the centerline self field. I changed the biot savart integral to do that, and also implemented an improved singularity-subtracted integral from the paper which should require fewer quad points, specifically for the compute_self_field function.
2. NaNs appearing in the unregularized biot savart integral were screwing with the gradient as well. I now have compute_self_field output a mask of where centerline points are, and the maxNormB objective function ignores those points (in a kinda gross way). I decided to do this over having a biot-savart safemode flag where it regularizes for points close to a centerline since the latter is more computationally expensive and might not play well with JAX JIT control flow logic. We don't expect peak field conditions to be anywhere in the vicinity of the coil centerline in any case.