Complete rewrite of DFT-D4 to get this project in a more sustainable shape. This change introduces API breakage (Fortran, C and Python API), therefore leads to a major version bump.
- improve unit testing and more testcases
- cleaner Fortran API via one single module
- more robust C API with `void*` based Fortran objects
- CFFI based Python API instead of ctypes based interface
- fixes issues with Intel 19
- uses [mctc-lib](https://github.com/grimme-lab/mctc-lib) as geometry backend
- remove EEQ implementation and create separate project
- uses [multicharge](https://github.com/grimme-lab/multicharge) library for EEQ model
- more stable choice of Ewald parameter (results for periodic calculations might differ slightly)