Non-blocking fetch implementation. (#83)

* Non-blocking fetch implementation.

* Make fetch tests more robust to timing issues.

* Bug fix when ignoring transferred objects.

* Fix.

* Documentation fixes.

src/plasma/lib/python/plasma.py

@@ -181,6 +181,14 @@ class PlasmaClient(object):
     """
     return libplasma.fetch(self.conn, object_ids)
 
+  def fetch2(self, object_ids):
+    """Fetch the objects with the given IDs from other plasma manager instances.
+
+    Args:
+      object_ids (List[str]): A list of strings used to identify the objects.
+    """
+    return libplasma.fetch2(self.conn, object_ids)
+
   def wait(self, object_ids, timeout=PLASMA_WAIT_TIMEOUT, num_returns=1):
     """Wait until num_returns objects in object_ids are ready.