extract some changes from pull request #504 that seem straight forward and helpful to speed up the travis builds

via distutils comment directives in a Cython source file.

when they are provided as commented distutils directives.

Add setup.py openMP instructions for Microsoft Visual C++ compiler

writing output file to a specified location.

follow PEP492 change in Py3.5.2 that makes __aiter__() a simple function instead of an async function
See https://bugs.python.org/issue27243

This posed an issue when the type parameter was a typedef.