Provide a collaboration graph / dataset

Rough idea: build the bipartite graph of who contributed to which project

Naive implementation: go from each origin (or last visit), and collect every author it references. But there is a lot of duplicate work.

Slightly less naive implementation: reverse-traversal in swh-graph, by tagging every revision/release/snapshot with the set of authors who contributed to their parent, starting from oldest revisions.

This might overflow memory though? If not, could we get away with a database?

---

Migrated from T4695 (view on Phabricator)