Enhance DC Inference to add basic AGM

[kylebd99]

Motivation: We currently have no way of taking cycles into account properly. For example, we can't recover the N^1.5 bound on triangles. However, we can recover the N^2 bound on 4-cycles, so it's not that bad.

I'd rather avoid using an explicit linear programming approach here if possible. One way to do this would be to add an explicit inference rule for triangles of DCs. This would probably handle the vast majority of cases here.