Probeye

Changed
- Small fix of a wrong type-specification in forward_model.py

Changed
- The forward model's init method can now take positional and keyword arguments.
- Small correction in docs (Mathematics sections, sigma_m -> sigma_m^2).

Changed
- Ensured reproducibility of solver results via seeds. It is also possible now to not use a seed (instead of a default seed). The reproducibility of the sampling results is explicitly checked in the unit tests of the solvers.

Changed
- Improved plotting functions to work better with a larger number of parameters.

Changed
- The definition of the latent parameter's priors was revised (from tuples to objects).
- Constant parameters are now added using a 'value' instead of the 'const' keyword argument in the InverseProblem's method 'add_parameter'.
- The forward model's response method is stripped from the problem definition so that the pure definition of an inverse problem indeed does not contain any computing methods.
- After parameters one must now define experiments and then the forward models.
- The four main attributes of InverseProblem are not private anymore to allow for manipulations in the solver.
- The solver routines were revised.
- Fixed some problems with the plotting routines.
Added
- Distribution classes have been added to make the prior-definition more clear.
- A new experiment class was added.
- A new correlation_model class was added.

Changed
- The likelihood function in the emcee-solver has been equipped with a check if the log-prior was evaluated to minus infinity to prevent unnecessary likelihood evaluations.