Test processing distribution on a multicore system using
``Py_BEGIN_ALLOW_THREADS`` and ``threading.Thread``.

Build with::

  $ python setup.py build

Test the time with::

  $ time PYTHONPATH=build/lib.linux-x86_64-3.2/ ./grab-cores.py 2

  real    0m10.075s
  user    0m18.364s
  sys     0m0.029s

You may need to tweak the ``PYTHONPATH`` to match the build directory
for your system.  The results shown are for my two-core system.
During the run, ``top`` showed the process pegged around 195% of the
CPU, so we are actually using all the available horsepower.