Centralize logic for graph I/O

[dorranh]

As a part of the changes in #39, a new module called explorer.graph was added which defines our core data model and key operations for interacting with the graph. As a matter of organization, it would be nice to have the etl process utilize this module to ensure that we are working with the same data model, etc. This will require us to add some additional data types to graph for elements such as edges and port functions for writing edges to the graph over to this module.

[GuillaumeDesforges]

Superseeded by #56, #41 and #65