Power-up the xpcs fitting function

xpcs.fit lacks a few more options for fitting.

As model, we currently have only a single exponential. A stretched exponential (KWW) should be added.

Fitting beta can be chosen as global (same for all q values) or local (independent) for all q. We saw in a couple of experiments, that also an analytical solution (with q dependence) can make sense. Should the third option for beta be analytical?