Continuum subtract arc spectra in 2-d

The HgAr lamp spectra that we are taking at LAM (and maybe all HgAr spectra!) have what looks like a significant continuum.

While not important for wavelength calibrations, this matters for PSF determination.

We think that all the spectra show this, but for definiteness:

Neven has looked at 8603, 8604, 8605 (all in focus), are 859* are also in focus, but dithered, and for instance for defocus is 8564.

The way to fix this is to do an extraction, mask the lines and fit the continuum, then use something like the self.debugInfo.residuals_frame code in reduceArcTask.py (the workhorse is ft.getReconstructed2DSpectrum(spec))