Optimise the pfsMerged/pfsObject wavelength scale

We're currently using a wavelength spacing of 0.081 nm/pixel for the pfsMerged and pfsObject products. This is in contrast to the blue dispersion of 0.066 nm/pixel and red dispersion of 0.088 nm/pixel. If linear, the wavelength spacing should probably be smaller than the smallest spacing anywhere in the instrument. Other, more complicated wavelength sampling schemes are possible (e.g., take the smallest instrumental wavelength spacing in any wavelength region, so the sampling would vary, sometimes even in a discontinuous manner).

We also need to decide how to handle the wavelength sampling when we're using the mid-resolution red grating.