Using a bipartite matching algorithm

[jmlarson1]

If this graph is always bipartite (which I think is the case) then this matching algorithm should be faster.

Timing for me on the following case:

(n, k): (288, 12)
orders: {x: 12, y: 12}
poly_a: x**2/y**3 + x + 1
poly_b: x**3*y**2 + y + 1

was about 15% faster... so if such a change isn't significant enough, then feel free to close this PR.

[perlinm]

Looks good to me, thanks!