The positivep attribute is degenerate, calculate on the fly

[karalekas]

As I was doing some debugging this week, one of the final things I had to fiddle with before inner blossom expansion started working was the positivep attribute -- even when the tree structure was set up correctly, an incorrect setting for positivep would throw off future recommendations and result in invalid states such as a node having two mates. A node's positivity is entirely determined by local information about the tree it is participating in -- for example, a bare vertex is positive, a vertex with a mate but no master is neither positive nor negative, a vertex with a mate that is also its master is positive, and a vertex with a mate that is also its slave is negative. Thus, there is no need to store an additional boolean positivep -- it opens us up to more possibilities for errors during blossom operations, and it can be calculated on the fly.

[ecpeterson]

I've become bearish on this idea. It's true that this information is redundant, but I'm not convinced it's any simpler to recalculate it on the fly. Also, this reasoning is opposite to what appears in #22 , also about tracking global tree state on participating nodes.