Reject bad FLUXSTD from flux calibration

We use F-type stars for FLUXSTD.  Stars with slightly higher/lower temperature are acceptable but they have to be within 5500<Teff [K]<8000 covered by the AMBRE library.  In Nov 2022 and Apr 2023 engineering runs, FLUXSTD included cooler stars and we need to reject them from the flux calibration vector. We need to think about how best we do that (e.g., these cool stars have bad chi-squares from the broad-band fitting and we could apply a cut there).

The attached plot is for visit=92114 (DA white dwarf).  The points show the parameter range covered by the library (projected onto Teff and logg), and the color-coding shows delta_chi^2<1, 2, 3 from the broad-band fitting. The star is the best-fit model.  The plot shows all FLUXSTD in that visit. As can be seen, we observed low temperature stars. The g-r color in the title is informative; 0<g-r<0.35 is about the range for F-stars.  Early G-stars will probably be useful, but we observed even K-type stars.