Telluric model fit performed in heliocentric frame instead of geocentric

[freddavies]

As titled. For spectra with logarithmic (dv or dloglam) pixels this can be fully accounted for by the fitted shift, but this could potentially cause subtle problems in linear (dlam) pixel space over wide wavelength ranges. Worth documenting somewhere that the presence of a large fitted shift of the telluric model does not necessarily indicate anything wrong with the wavelength solution...