KPFM is wrong because of sign issues

[mondracek]

The sign change of the electrostatic potential and charge density, implemented in PR #124, was not followed in the KPFM part of the code. As a result, sign convention in the KPFM part was inconsistent (some potentials being expressed in V, inherited from the main part of the code, others remained in eV; likewise for densities that could be expressed as either charge or electron densities, making a sign difference). KPFM was giving nonsensical results because of this.

Details shall be obvious from the code changes in a Pull Request I am about to create to address this issue. This issue needs to be settled before it makes any sense to go on with #243 (or even #244, #245).