1) Better radial distribution stats by 'randomizing' the selection
Does the radial distribution select periodically from all the chains?
What I mean is, if you set the "analyzed for radial" to 1e7 with the
total number of chains set to 1e8, are the FIRST 1e7 analyzed or are
the modulus10 analysed?

It should be modulus10.  Otherwise, you will preference certain half-chains
in the result.  Defeating the whole purpose of the independent 
calculations idea.

2) Replace 'hpsort' in math_routines.c with a GPL version.
