explicit comparison function case:  use PyObject_Call instead of
PyEval_CallObject.  Same thing in context, but gives a 2.4% overall
speedup when sorting a list of ints via list.sort(__builtin__.cmp).