Explore Wavefront tracking the phasor as the fundamental quantity #295
Description
Per #291 nans can arise when subtracting wavefront phasors (necessary for fast computation of Lyot systems, Zernike WFS, Pyramids, etc).
This issue arises when casting from complex (Cartesian) back to Polar, in the np.angle call since the denominator of the operation is zero. This can be fixed with a np.where call but its inelegant and can break power conservation. If all operations stay on the complex phasor then this issue is circumvented entirely.
The original decision to track amplitude/phase was to allow for easier interpolation of wavefront phase (a rarely if-ever used capability). @ataras2 suggested a custom combined amplitude/complex interpolation routine that may circumvent this issue.