Final feature scaling

Created by: RoyStegeman

This PR adds the option to set the variable interpolation_points in the fitting dictionary of the runcard.

Activating this option will change the architecture from x^alpha*(1-x)^beta*NN(x,logx) to x^alpha*(NN(scalex(x)) - NN(x=1)), where scalex is a function which maps the xgrid->scalex(x) in such a way that the x-values of the input FK tables are almost homogeneously distributed on the domain [-1,1]. The 'alsmost' in the previous sentence refers to the fact that after performing this transformation such that the new input points are exactly homogeneously distributed, a number of functions equal to interpolation_points are used to interpolate between an evenly separeted subset of input points. This is done (as opposed to just interpolating between each two neigboring input poits) to include information on the interpolation in the input data.