Draft: construct pangenome database from GFA

Related to !147, I would like to not only go from pantools_DB -> GFA but also vice versa. This merge request collects my tries of doing that. It's marked as draft for now because I'm not sure if I'll use this code in the end.

TODO:

• Fix kmer index
• Make working on cyclic graphs
• Find out what happens with degenerate nodes
• Don't require genome locations file if GFA is provided (if GFA contains sequences)
• Get it to run on any valid GFA
• Ensure neo4j -> GFA -> neo4j gives identical database
• Ensure GFA -> neo4j -> GFA gives identical GFA file
• Update documentation accordingly

docs/source/construction/group.rst

@@ -38,7 +38,10 @@ functions.
   given threshold (controlled by ``--relaxation``), the proteins are
   connected with each other in the similarity graph. Every similarity
   component is then passed to the MCL (Markov clustering) algorithm to
-  be possibly broken into several homology groups.
+  be possibly broken into several homology groups. MCL can create
+  overlapping groups if two proteins within a component have identical
+  similarity scores, in this case the overlapping groups are merged and
+  labelled as a "merged" group.
 
 | **Relaxation**
 | The ``relaxation`` parameter is a combination of four sub-parameters: