TODO

From a training session w/ Julian Christou and Chad Trujillo:

- remove need for zeroes.fits (worked around)
- output coefs
- output phase
- output images fits
- display image residual
- output ncpa in correct format for altair
- pupil option (user pupil, best is to pass name of pupil generation function)
- output dh in terms of Zernike (fit phase map with zernikes)
- solve a few bugs:
 - progressive = 1 crashes when nmodesmax<45
 - check unit of kernel is FWHM, not rms
 - clean up input parameters (started)
 - why internsity is not centered on 1.0? because of clipping in data. 
   option to clip display from zero?
- include effect of pixel averaging (noticeable if fwhm close to 2).