Clustering - rework the shadow matrix

[GoogleCodeExporter]

The shadow matrix is used in clustering as a final combination of all members 
across all matrices used in clustering when more than one matrix is specified.  
This means that all possible pairs are considered when more than one spatial 
condition is specified.  

Currently it requires special handling in the code.  

A cleaner approach would be to add an extra matrix to the list of matrices in 
the build phase.  This will then be handled by the cluster algorithm.  

There are two possible approaches.  The first is to add a spatial condition 
which is the union of the preceding spatial conditions.  The second is to get 
the list of matrix elements across all existing matrices once the first set are 
calculated, and then build a matrix containing those pairs not already in the 
other matrices.  

The former approach will need consideration of spatial optimisations, while the 
latter needs to check existence in preceding matrices.  

This will also be a reworking of the shadow matrix approach, as currently it 
stores the union of the existing matrices.  

Original issue reported on code.google.com by shawnlaffan on 3 Jun 2014 at 11:08