[EM] Median bifurcation using pairwise matrix

[Roy]

I'm afraid I've scrambled two different methods under one name. 
Median bifurcation, as I originally intended it, groups the 
candidates based on median *rank*. As Forest pointed out, a majority 
of voters can rank the CW in the lower half, which would rule him 
out, so it fails Condorcet (but it takes a pretty polarized 
electorate to do that).

The pairwise matrix version doesn't use median rank, it uses wins vs. 
losses, which may be some kind of median, but it's not an equivalent 
method. I recommend calling it Win-Loss bifurcation; it is (as Rob 
LeGrand pointed out) very similar to Nanson's. In fact, it's probably 
equivalent: the only difference is that mine uses Row_total - 
Column_total and compares to zero (which is average), whereas Nanson 
compares row totals to the average of the row totals. It's the first 
example I've seen of bifurcation in election methods. But then, I'm 
not particularly well-read yet.