Fix "PyErr_Format + PY_FORMAT_SSIZE_T likely broken in Win64"

Add '/openmp' flag when compiling Cpp code with msvc compilers

refactoring and rewrite of optional kw arguments parsing helper function to support Unicode keyword arguments in Py2 (CPython supports them in Py2.6+)

Fix parallel/prange test failures on Windows

move utility code inclusion from Name/AttributeNodes to the latest possible step in the pipeline (after removing it from declaration analysis)

always instantiate exceptions directly when raising them, fix Py3 crash when an exception type does not create an exception instance (works in Py2, although not in PyPy)

prevent utility code of overloaded function entries from accidentally being written to the output file

added safety assertion to prevent utility code from overwriting that of other entries with the same name

use proper module.m_free() way to run module cleanup code in Py3, way after running atexit functions etc.

moved utility code for registering the module cleanup function into Cython/Utility/ModuleSetupCode.c