This branch basically was another attempt at fixing the potential fitting.
In EmpiricalPotential_contain_HomologyGraph we settled with simply knowing this
homologous graph representing the binding model of the respective potential,
e.g. star-like with three aroms for improper, chain of 4 for torsion, ...
Here, we extracted the edge information from the argument_t, i.e. the atomic
number and indices at either end. Also, we added the clas Subgraph that should
contain the potential's binding model. With this, we eventually wanted to be
able to set a bonded flag in argumen_t (as on the other branch).
This branch can be removed when EmpiricalPotential_contain_HomologyGraph seems
to work.