Clustering - accelerate the tie breakers

[GoogleCodeExporter]

What steps will reproduce the problem?
1.  Run a cluster analysis with a large matrix which contains many zero values, 
e.g. using a Simpson or phyloSimpson index
2.  The tie breakers take a long time to be evaluated for the first round.  
this is simply due to the large numbers of ties.


This is working as designed, but in many cases there are groups of matrix pairs 
which ultimately are collapsed into a single cluster.

It might be possible to search for symmetric blocks of the matrix which can be 
merged in one step, thus skipping the tie breaker calculations.  

Original issue reported on code.google.com by shawnlaffan on 12 Jan 2014 at 7:03