PyBigDFT Improvements

1. Pseudopotentials converted to a YAML format Note that the old string format will still work. Also: the files are stored with a yaml extension, so that people don't get confused and try manually copying them into the run directory.
2. Fixed verbosity of calculator
3. Database naming - switched from comma separated to -_- separated. Maybe we can make , work on all file systems, but this seems like the easier solution.
4. Inclusion of Krack pseudopotentials in the database.
5. New protocol for remote runner which uses jsonpickle. This is a pure text format, so it might be safer with exotic systems.