Lsqfit

Tiny fix to setup.py so lsqfit installs properly when the GSL library is not present. (Note that lsqfit is more useful with GSL than without it, but does work without it.)

This version offers a major overhaul of the underlying fitting strategies used by `lsqfit`, featuring more modern alternatives that can be more robust and/or much faster than before. See the Faster Fitters section of the tutorial in the documentation for details. Large parts of the documentation have been rewritten, especially in the tutorial which has more and simpler examples. New options exist for fitting very large uncorrelated data sets. This is the first version of `lsqfit` that can be created without Gnu's GSL library (provided `scipy` is installed), although building with GSL is still recommended.

Very small change to better integrate MPI support for lsqfit.BayesIntegrator.

Minor fixes for compatibility with gvar v8.0.

Small change to documentation but otherwise same as v8.0

New class BayesIntegrator for evaluating Bayesian integrals. New parameter distributions.