emission lines are affecting the normalization of the BGS, ELG templates

[moustakas]

The ELG and BGS template classes -- which have emission-line spectra added to their underlying continuum spectra on-the-fly -- are not being normalized correctly. For example, if the user requests a set of BGS spectra all normalized to r=19.5, the output r-band magnitudes of the templates varies between r~19.44 and ==19.5, with the residuals correlating with the equivalent width of H-beta. This happens because I'm normalizing the spectrum before adding in the emission-line spectrum (for non-trivial reasons). Need to fix this.