think more about different PSF centering between Fresnel & Fraunhofer modes

[mperrin]

For historical reasons related to sampling choices in MFT vs. FFT algorithms, output PSF centerings are slightly different between Fraunhofer & Fresnel calculations for even number of pixels. Here I'm talking about where the PSF peak intensity is, in the ideal case with zero tilt etc.

• Fraunhofer mode: PSF is centered in the exact middle of the output array, e.g. for 128^2 pixels, the PSF is centered at (63.5, 63.5), on the corner in between the 4 central pixels.
• Fresnel mode: PSF is centered on (64, 64), on the integer pixel closest to the center but above it. (The 0,0 pixel after a fftshift operation)

This is fine enough but (a) the behavior should be documented more clearly, and (b) Various utility functions for display, radial profile computation, etc. should perhaps be updated to use the correct center for Fresnel mode PSFs. (c) Perhaps we should add some FITS keywords to output PSFs indicating where the correct center reference pixel is?

Overall this more like a subtle semantic issue, not an actual problem with the output PSF values.