Trace shifts only in python and not in specex

[julienguy]

So far, the trace coordinates and wavelength solution have been refined in the specex PSF fit.

This is in principle more precise than a cross-correlation method or a simple Gaussian fit of emission lines because it is a full forward model of the CCD image, with a more precise PSF, integrated in the pixels. However, it is CPU intensive, it's a non-linear fit so the convergence time fluctuates, and it can fail in case of unmasked cosmic ray hits on top of some emission lines.

The alternative (trace coordinates and wavelength solution only in the python code, with the already existing trace_shift code) should be evaluated.