Parse start->action, attribute->end counts out of DBIR data

[gdbassett]

Because start/end edges are so common in path mitigations. they should be parsed out of the data.

It should be easy to identify when an action is the first one or an attribute is the last one and increment counts on the edge from/to the start/end nodes.