Issue #25221: Fix corrupted result from PyLong_FromLong(0) when Python is compiled with NSMALLPOSINTS = 0.

between 1.0 and 2.0 to 1.

generalizations added in 3.5.

This is required by RFC 1542, so despite the subtle behavior change we
are treating it as a bug.  Patch by Michael Lazar.

This is now an official "protected" API that can be used to write
classes that are duck-type-compatible with Future without subclassing
it.  (For that purpose I also changed isinstance(result, Future) to
check for this attribute instead.)

Hopefully Amber Brown can use this to make Twisted.Deferred compatible
with asyncio.Future.

Tests and docs are TBD.

issue14976.

This is buggy in the old email code as well, but it doesn't affect anything
there because only the default line length is ever used there.

Too many build tool chains do not properly support it, including building
and linking an executable fine that simply segfaults when you try to run
it (such as debian jessie 8.5's gcc 4.9).  On others where it does appear
to build (ubuntu 16.04's gcc 5.4) there are still test_distutils and test_gdb
failures to deal with.

We're not going to spend time attempting to maintain a complicated white list
of what does and doesn't work in our configure.ac file.