suppress flux up/down-turn in the n-arm

In the current version of fluxCal, we do not let the shape of the flux calibration vector vary over the focal plane. But, we do see a spatial structure in the color difference between PSF and PS1, suggesting that we do need to let the shape of the flux calibration vector vary to correct for it. However, the reddest filter available is PS1 y, and the flux calibration vector in the n-arm is essentially an extrapolation. This extrapolation is unstable and often results in artificial flux up/down-turn as shown in the figure here (it is pfsSingle of a FLUXSTD). One way to suppress such an artificial feature is to add a data point beyond the PS1 y-band, so that we can sort of regulate the extrapolation. This ticket explores how well it works.