Parrellize cythoncpp.py.

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.)

gevent/callbacks.c

 /* Copyright (c) 2011-2012 Denis Bilenko. See LICENSE for details. */
 #ifdef Py_PYTHON_H
 
+#define _GEVENTLOOP struct __pyx_vtabstruct_8corecext_loop
+
 static void gevent_handle_error(struct PyGeventLoopObject* loop, PyObject* context) {
     PyThreadState *tstate;
     PyObject *type, *value, *traceback, *result;
@@ -19,7 +21,7 @@ static void gevent_handle_error(struct PyGeventLoopObject* loop, PyObject* conte
 
     PyErr_Clear();
 
-    result = ((struct __pyx_vtabstruct_6gevent_8corecext_loop *)loop->__pyx_vtab)->handle_error(loop, context, type, value, traceback, 0);
+    result = ((_GEVENTLOOP *)loop->__pyx_vtab)->handle_error(loop, context, type, value, traceback, 0);
 
     if (result) {
         Py_DECREF(result);
@@ -195,7 +197,7 @@ static void gevent_run_callbacks(struct ev_loop *_loop, void *watcher, int reven
     loop = GET_OBJECT(PyGeventLoopObject, watcher, _prepare);
     Py_INCREF(loop);
     gevent_check_signals(loop);
-    result = ((struct __pyx_vtabstruct_6gevent_8corecext_loop *)loop->__pyx_vtab)->_run_callbacks(loop);
+    result = ((_GEVENTLOOP *)loop->__pyx_vtab)->_run_callbacks(loop);
     if (result) {
         Py_DECREF(result);
     }

greentest/testrunner.py

@@ -9,12 +9,13 @@ import time
 from datetime import timedelta
 
 from multiprocessing.pool import ThreadPool
+from multiprocessing import cpu_count
 import util
 from util import log
 
 
 TIMEOUT = 180
-NWORKERS = int(os.environ.get('NWORKERS') or 4)
+NWORKERS = int(os.environ.get('NWORKERS') or max(cpu_count() - 1, 4))
 
 
 # tests that don't do well when run on busy box