#120 Make Redfield functionality usable + more updates

[Ulf Lorenz]

In ticket [#97], some refactoring work has been done on a RedfieldFactory to simplify the solution of Redfield problems. However, a problem that cropped up was that the current RedfieldFactory requires you to know the correlation function of the bath.

In most cases, you do not have a correlation function, but a spectral density of a harmonic bath, and the correlation function cannot be calculated in closed form. Hence, you want to have a setup where you supply a spectral density and get the Redfield problem set up. A reference here is Pollard and Friesner, JCP 100:5054.

Implement the necessary functionality so that this can be easily done. Preferably start from a demo and implement the functionality that seems recurring work.

Other notes:

• you may also need functionality to check for convergence easily.
• as long as you do not have absorbing boundary conditions, the Redfield operators Lambda_m are self-adjoint. Does it make sense to add a simpler interface that requires only one operator, instead of Lambda_m and its adjoint?
• add a useful Redfield demo and remove the thereby superseeded"Redfield" acceptance test
• do not forget the Python interface; currently, the Redfield functionality is not tested