Poor-quality IQ measurements for longslit spectra with auto-found apertures

[chris-simpson]

When executing measureIQ() on a longslit spectrum with apertures, multiple measurements are made for each aperture (at a range of wavelength slices, so the measurement is not broadened by s-distortion), which are then combined as if they are all independent measurements. Unfortunately, if one or more aperture is false (as often happens in automated finding) it's hard to reject these and they can skew the result.

Some smarter logic is probably required that, for example, compares the multiple FWHM measurements of each source and rejects all of them if they're inconsistent.