Re-enable the RLock threading tests on Python2.

They seem stable now.

src/gevent/_abstract_linkable.py

@@ -324,6 +324,10 @@ class AbstractLinkable(object):
         self._check_and_notify()
         # If we added unswitched greenlets, however, don't add them back to the links yet.
         # We wouldn't be able to call them in this hub anyway.
+        # TODO: Instead of just adding these back to self._links, we should try to detect their
+        # "home" hub and mode the callback to that hub. As it stands, there's a chance that
+        # if no greenlet tries to acquire/release this object in that hub, these objects
+        # will never get to run.
         self._links.extend(unswitched)
 
     def _quiet_unlink_all(self, obj):

src/gevent/tests/test__lock.py

@@ -11,9 +11,6 @@ from gevent.tests import test__semaphore
 
 class TestRLockMultiThread(test__semaphore.TestSemaphoreMultiThread):
 
-    # For some reason, the tests are extremely flaky on Python 2.
-    dueling_thread_tests_are_too_flaky = greentest.PY2
-
     def _makeOne(self):
         # If we don't set the hub before returning,
         # there's a potential race condition, if the implementation

src/gevent/tests/test__semaphore.py

@@ -238,14 +238,9 @@ class TestSemaphoreMultiThread(greentest.TestCase):
     def assertOneHasNoHub(self, sem):
         self.assertIsNone(sem.hub, sem)
 
-    dueling_thread_tests_are_too_flaky = False
-
     def test_dueling_threads(self, acquire_args=(), create_hub=None):
         # pylint:disable=too-many-locals,too-many-statements
 
-        if self.dueling_thread_tests_are_too_flaky:
-            self.skipTest("Described as too flaky")
-
         # Threads doing nothing but acquiring and releasing locks, without
         # having any other greenlets to switch to.
         # https://github.com/gevent/gevent/issues/1698