The profiler filters lines on the function's __code__ attribute, which
must agree with that created in the profile frame.  Unfortunately, this
makes it difficult to profile anything but 'def' functions.

always create a new IntNode for the step value in iter-range optimisation as we overwrite its value later

Optimized reversed range with constant negative steps.

extend reversed iteration tests, including disabled tests for negative step values that fail when enabled

--HG--
extra : transplant_source : %F9%BB%B7%FD%A5jc%89%02%BD%BANie%A3%FDc%8F%9Ct

Use PySequence_List, PySequence_Tuple

Cython array import naming conflict

do not depend on reversed(range()) bounds having a numeric type and just use whatever spanning type we find

Optimized reversed-range call when constants are inside of the range.

postpone import of distutils Extension type in test runner in case setuptools gets imported in between

postpone import of distutils Extension type in test runner in case setuptools gets imported in between

--HG--
extra : transplant_source : %9B%C2%08Q%DDt%9A%99%A4%CAK%5D%8F%3A%A9%E4F%19%96%CD

fix test runner and pyximport in Py 2.7.9 and 3.4+: setuptools replaces "distutils.dist.Distribution" *after* we import it, which then breaks the input validation in "distutils.cmd"

fix test runner and pyximport in Py 2.7.9 and 3.4+: setuptools replaces "distutils.dist.Distribution" *after* we import it, which then breaks the input validation in "distutils.cmd"

--HG--
extra : transplant_source : %3A%84%AC%02E%17%D8%EE%E64Mpv%25%A3%8E%8B%E5%CA%19

Minor cleanup (2.5 compat, mostly)