Improved numerical derivative

[oliverhaas]

For the calculation of the numerical derivative for the backward Abel transform I rely on finite difference. Since the user inputs equidistant data this is fairly natural and works pretty well overall. In general there might be better ways to calculate the derivative numerically (look for Cauchy Integral derivative or at the chebfun Matlab package).
I don't see much value in it now (in my work I can almost always take the derivative analytically), but if there arises a use case one could look into this.