Pass seed groups directly to SybilGroupRank rather than finding seed groups by looking at seed users

[abramsymons]

https://github.com/BrightID/BrightID-AntiSybil/blob/8ebe6ccb407d487ce4f060976dedf71c8d137d4e/anti_sybil/algorithms/sybil_group_rank.py#L36

If we have for example 3 seed groups A, B, C and A=(a1, a2, a3) B=(b1, b2, b3) and C=(c1, c2, c3) and then (a1, b1, c1) make a group, my approach set this new group as a seed group. It may be problematic.

There is a subtle attack where members of different seed groups can choose to collude to create new seed groups. It doesn't seem likely, but it's probably better to avoid it.

[adamstallard]

To sum up the bug, if all members of a group are in seed groups, that group will be a seed group even if it wasn't designated as one.