Quick workaround for ray.wait bug.

[robertnishihara]

Without this fix, you can reproduce the bug as follows (on Mac OS X).

import ray
ray.init(start_ray_local=True, num_workers=1)

@ray.remote
def f():
  return 1

ray.wait([f.remote()], timeout=(2 ** 60))

The last line just hangs.

I've looked through the code, and the timeout is properly passed around as a 64 bit int (all the way to where the timer is added to the event loop). The hanging occurs because when the result of the task f is written to the object store, it calls seal_object in the plasma store (like it's supposed to), which calls send_notifications to notify the plasma manager (like it is supposed to), but the corresponding callback in the plasma manager (process_object_notification) never gets called (and that's the part that should cause wait to return).