Required a small amount of object changes, and then a large
number of replacements where we directly compared the class
to int_cls.  We now check that it's a subclass, since you
are allowed to use True/False anywhere that you could use
an int.  ex "xrange(0, n, step=True)"

in both directions, from Pyston->CAPI and CAPI->Pyston

Otherwise those symbols don't end up getting exported and things
crash if we try to do any inlining, which only happens on the larger
benchmarks or when running in -O mode.

travis run full ninja check-pyston and update README

As much as I don't like modifying the standard libraries,
removing the dependence on this particular one helps a lot,
since it (indirectly) imports weakref which will take a while
for us to support.

ie by running 'make check USE_CMAKE=1'

changed:
- defer to cmake for building the unittests
- building extensions doesn't use the rest of the build system so can remove
  that unnecessary dependency (would cause an llvm build otherwise).

Should also reduce our Linux-dependence

I was surprised that Py_BEGIN_ALLOW_THREADS was working, but it turns
out that we weren't enabling threading so it wasn't doing anything.

Compiling posixmodule.c under ccache in certain systems ends up
causing trouble.

fixes #4 add travis-ci.org support

Cmake add pwdmodule, slots_test, and match existing test configurations

fixes #238 CMake issues

We weren't able to exit if we had a thread running in the background,
since we released the GIL at the end of the C level main() and the
background thread would start running again after all the data structures
had been torn down.

Now, just keep holding the GIL as we exit.

Was tired of seeing bare "unimplemented" error messages and having
to track down where they were coming from -- especially tricky
when trying to add fork() support and the errors were in the child.

- int(bool) works
- hash(bool) works
- str.find() now takes an optional "start" parameter