fix statistical data merging logic

[eiennohito]

This PR makes statistical data merge logic correctly merge transitive and partially intersecting sets.

See tests.

Notation: pre-merged sets are in brackets, individual elements are noted with letters.

Now situations like:

• (A, B, C), (B, C, D), (D, E, F) -> (A, B, C, D, E , F)
• (A, G, H), (B, F, H), (C, H, X) -> (A, B, C, G, H, X)
  are handled correctly.

Also fixes the --no-ones flag which had its meaning inversed :|

Also bumps spark/scala versions.

@mh-northlander you may want to see this