Mismatched trace regions and wavelength regions lead to incorrect wavelengths

[gmbrandt]

There was a bug where slight ~1 pixel noise in the trace centers led to extracted wavelengths being 10/11 's what they should be. We fixed this in PR #74 . But this was a temporary fix that basically takes the center pixel's wavelength along each column. If we ever go to a fully 2d wavelength solution (and non vertical extractions) we will need to apply a more robust fix (e.g., always using the same super lampflats for both the wavelength solution and extractions).