Rewritten response code with improvements

The overarching goal is to make the response code more accessible. From experience, it has proven to be difficult to implement new features without a overview of the entire code.

At the moment, chiks.py shows how a specific type of Kohn-Sham linear response function can be implemented in a straight-forward way from a more general base class.

Only one type of interacting linear response function, found in susceptibility.py, has been implemented, so this part has not been generalized. The FourComponentSusceptibilityTensor of susceptibility.py has an interface for the transverse magnetic part in tms.py, but an interface for the dielectric function is left for future work.

While rewriting the code a number of improvements have been made. The code as a whole should be faster due to smarter summation of transitions between Kohn-Sham states and improved vectorization. Furthermore, a memory-distributed ground state calculator object can be used as input now, thus avoiding having to write wave functions to disk.