rename "gi_running" attribute of coroutine type to "cr_running" and adapt some docstrings to the type split

re-implement Coroutine.__await__() using a dedicated wrapper type to avoid having to make a Coroutine half an Iterator

Fix CFLAGS environ in tox config

Add signbit to math.pxd.

Also reorder a few imports to alphabetical order.

Conflicts:
	Cython/Shadow.py

keep C++ code out of C files and vice-versa by splitting module preamble (original patch by Lars Mans)

exclude two more modules from stdlib compilation that require bootstrapping before importing Cython modules

Write "generated by Cython" header also in .h files