[EM] Median bifurcation using pairwise matrix

[Roy]

> FYI, I believe this procedure will give the same results as the
> procedure called Nanson 

Actually, no. Using Forest's recent, very sticky example, we get a
Borda matrix of

   A   B   C   D   X
A  -   71  42  29  39 
B  29  -   71  58  39
C  58  29  -   87  39
D  71  42  13  -   39
X  61  61  61  61  -

which yields, by Nanson, X>A>B>C>D because D is eliminated first

but by my proposed method, X>C>D>A>B because A, B and D are eliminated
(or grouped) together, then A and B are eliminated from that group.

(and by another grouping method I'm considering: X>B>C>D>A)

But when considering methods, you have to ask WWRPD (What Would Ranked
Pairs Do?)

The winning-margins matrix is:

   A   B   C   D   X
A  -   42  -   -   -
B  -   -   42  16  -
C  16  -   -   74  -
D  42  -   -   -   -
X  22  22  22  22  -

With all the duplicate values, I get confused. WWRPD?