Preprocessing of topology

[orlandoisepic]

Short description

Currently, there is some preprocessing done by both TopologyReader (in preprocess_participant_names() and check_topology()) and NodeCreator (in _data_preprocessing(...) and _patch_preprocessing()).

These methods update the topology-dict (derived from the topology.yaml-file) and thus need to be called after checking valid syntax but before doing the actual node generation.

However, the preprocessing in NodeCreator already requires some form of knowledge about the topology, namely the participants the info contained by their nodes (name and dimensionality).

If the preprocessing is split into a different step, the participants would need to be created / handled outside of the NodeCreator, which also not so nice.

** Proposed solution **

Do the preprocessing before and see if we can do the participant thing twice without causing too much troubles.

[orlandoisepic]

After #15 , a lot of the preprocessing in NodeCreator is moved to TopologyReader.
The last preprocessing in the creator is _data_preprocessing() which can also move to the reader, as it does not really require any information about the nodes themselves.

The topology reader thus has a lot of preprocessing.
First, it should be simplified a bit by making it modular, i.e., the long loop should be split into smaller ones.
Next, it currently does this in a method check_topology(). This method should thus only include checks, not transformations itself. These should be moved into a separate method.

[orlandoisepic]

The _data_preprocessing() method is redundant with the changes proposed in #17. It only "uniquified" data (did some preprocessing steps) if exchanges were unique up to the "from-/to-location".

As it is now allowed to do a list in "from-/to-location", the method is counterproductive.

With this, all preprocessing happens in "Topology reader"