busclique finds cliques that aren't confirmed by the python polytime embedder

[boothby]

edit: There's an issue in how the polytime embedder handles bad internal couplers

See #151 and #152. Seems to be impacting chimera graphs with a small number of bad edges

[boothby]

Oops. I misread those logs. The issue is actually in the python polytime embedder, which isn't (yet) minorminer's responsibility. Leaving this open because I'll eventually want to fix that.

[arcondello]

@boothby , potentially closed by dwavesystems/dwave-system#325?

[boothby]

Yes-ish? There's still some value in that python code:

1. it can handle arbitrary shore sizes; the c++ is limited to 8
2. it can be used to sample from the space of maximum-size embeddings (with caveats), where the c++ is deterministic (with caveats)

So for now, I don't plan to remove it altogether.