Fix a slight possibility of not aborting until after twice the reques…

…ted timeout.

Normally by the time self._results.get() times out the abort timer (self._timer) should have already
fired.  But there's a remote possibility that it hasn't, in which case the existing self._aborted.value
check returns False, and we continue the loop into another self._results.get() call.

Instead just make sure the abort timer has already fired by joining on it--in most cases this will
immediately return.

src/sage/parallel/map_reduce.py

@@ -1134,13 +1134,12 @@ def get_results(self, timeout=None):
                              (active_proc, timeout, self._aborted.value))
                 newres = self._results.get(timeout=timeout)
             except Empty:
-                aborted = self._aborted.value
-                logger.debug('Timed out waiting for results; aborted: %s' %
-                             aborted)
-                if aborted:
-                    return
-
-                continue
+                logger.debug('Timed out waiting for results; aborting')
+                # If we timed out here then the abort timer should have
+                # already fired, but just in case it didn't (or is in
+                # progress) wait for it to finish
+                self._timer.join()
+                return
 
             if newres is not None:
                 logger.debug("Got one result")