This code provides a comparison between a few varieties of funnel sort, and Tim Peters' timsort - the latter being
the standard sort included with Python.

Funnel sort fails to outperform timsort, at least when Funnel sort is implemented in Cython and uses few of the tricks
seen in timsort to go beyond a simple combination of merge sort and insertion sort.

However, interestingly, of the variants of funnel sort tested, it does indeed appear that using an exponent of 1/3
is best, as suggested by the theory.  That's a somewhat satisfying result, and suggests that perhaps timsort should
be modified to use funnelsort instead of merge sort in the large to get cache obliviousness.

It's currently (Wed Jun 23 12:52:12 PDT 2010) about 6.5 times slower than timsort, however looking at the Cython-
generated C code, it appears there's a lot of potential for speed improvements by writing a C extension module
manually, instead of autogenerating C from Cython.