Fix jellium code to allow it to work with the dipole correction

This entails a slight change in how the JelliumPoisson solvers are initialized. The charge (=ne) has to be supplied when instantiating it.

Jellium surface calculations with dipole correction converge relatively slowly (since the density starts out completely homogeneously), and that it might be worth to converge such calculations first without dipole correction and then add the dipole correction.