Move cdef and source from Python strings to .c files that can be parsed and validated by a real C compiler.

reorganize changelog entries for 1.2a1 to group related entries. it was getting hard to follow. [skip ci]

Actually it wasn't quite as dire as described there. Instead, there were
just extra copies of each chunk of data the application yielded.

This commit removes the copies, sending the data directly to the socket.
This should be most noticeable on larger chunks.

Also re-organize the call to close() the app iter closer to where the
app iter gets assigned, making it easier to spot. Add a test for
this---turns out the validator was closing the iterator for us!

I found one genuine cycle in callback functions. The rest I can't find
any legit leak, so for now I'm disabling that part of the test on the TPE.

It now works as expected when both patched and not patched. It is
cooperative with greenlets in both cases. Test this.

This can be helpful when threading is patched.

Ref #786.

I can't reproduce with Python 2.7 locally, and the error makes no
sense (because we're writing the banner to a *brand bloody new* file,
guaranteed not to exist). So this is a shot in the dark.

See
https://ci.appveyor.com/project/denik/gevent/build/1.0.568/job/p44icqoi91a72e08.

Failed to rename 'c:\users\appveyor\appdata\local\temp\1\tmp_j5icn\gevent\corecext.pyx.tmp.2832' to 'c:\users\appveyor\appdata\local\temp\1\tmp_j5icn\gevent\corecext.pyx
Exception in thread Thread-1:
Traceback (most recent call last):
  File "C:\Python27\lib\threading.py", line 801, in __bootstrap_inner
    self.run()
  File "util\cythonpp.py", line 31, in run
    self.value = target(*args)
  File "util\cythonpp.py", line 432, in _run_cython_on_file
    atomic_write(unique_pyx_filename, py_banner + value)
  File "util\cythonpp.py", line 902, in atomic_write
    os.rename(tmpname, filename)
WindowsError: [Error 32] The process cannot access the file because it is being used by another process

This permits some more optimizations and speeds up the build another 20%
or so, in addition to being hopefully easier to follow.

On my machine, this cuts the build time by a fourth to a third.

Not ready to merge, submitting for testing on CI (esp Windows).

Cython calls are threaded because they release the GIL.

The most expensive part was merging results, which holds the GIL, so use
multiprocessing for that. Pre-combine the identical results that we
expect to get for two sets of defines to reduce the number of merges.

Change internal data structures to be immutable to make debugging this
easier (and substantially faster when pickling through multiprocessing.)

Don't subclass dict for environ by default, but leave that option
available. When printing, make sure that we have a secure environ.
Unfortunately, this is relatively fragile but it's the only way to not
break WebOb/Pyramid.

Right now, this is only used for suppressing the printing of potentially
sensitive information, but in the future there could be other uses.

This is technically not compliant with PEP3333 which specifies that the
type(environ) must be dict, but it's not clear if that practically
matters anymore (it looks like it might be a holdover from supporting
Python 1.5.2 before one could subclass the builtin dict; see https://mail.python.org/pipermail/web-sig/2003-December/000394.html).

Fixes #779.

test__socket_close specifies AF_INET sockets so make sure and use an AF_INET address in the server. This affects only Windows.

- Additional testing on a mix of ip4 and ip6 connections.
- Simplify address parsing.
- changelog.

Fixes #495.

That'll make the socket work both with IPv4 and IPv6.

Comments inline.

[skip ci]

We now depend on CFFI/Python new enough that it should be built during
setup/installation.

Trying to build at runtime easily results in "gevent" directories
scattered all over the system.