Leave lamp response in the master flat

Spectra extracted from the image must have been divided through by the response of the system to the flat-field lamp; that is, we should not make any correction for the effects of the flat-field lamp or the dichroic in the per-arm spectra, and we should not correct the merged spectra for the flat-field lamp spectrum until we look at the spectrophotometric standards (at that point in open-use when no spectrophotometric standards have been observed we may want to do something special/hacky).

Currently, this is done by normalising the flat-fields row by row, and putting the lamp response in the fiberTrace. However, we want to reverse this: the flat should not be normalised row by row, and the fiberTrace should be normalised row by row. That flat should be normalised as a whole across the entire instrument, as is done for imaging.