Separate the parser and the neo4j backend

[alexgarel]

The idea is to refactor current parser so that we have:

• the parser class that just parse the file and yield the different nodes (with a type)
• a backend that iterates over a set of nodes in the format released by the parser and add them to the Neo4J database

There should be an abstract GenericBackend base class.

Checks should be implemented on the backend side if needed.