Use existing code for arclength

Created by: scarlehoff

I've redone #712 this is ready for review.

Not ready for merging because after/if people is happy with the way this is done I'll add checks and perform a fit to ensure that the arclenghts are correctly computed. But I don't think it makes sense to do that if the idea is terrible.

I've created a N3PDF class which basically hides validphys the fact that it is not dealing with a LHAPDF set or a nnpdflib object. As more and more validphys actions stop depending on the particulars of nnpdflib/lhapdf the N3PDF will be losing hacks or/and they will become features.

This is the only way I could think of start using features of validphys without rewriting half of the functions (because nnpdflib has roots down to the core).

Closes #558 (closed)

Fit comparing master and 08cce4d9 : https://vp.nnpdf.science/xVHH8R7mSHqVOXLRTZK73A==/