Fixed a bug with dropped connections and async requests.

When a connection is dropped, all "wait" operations of asynchronous calls
are now released with an error message, exactly like synchronous calls.:

sjrpc/core/rpcconnection.py

@@ -396,12 +396,16 @@ class RpcConnection(object):
             logging.debug('Error while socket close: %s.', err)
 
         # Release all running calls from this connection:
-        for call in self._calls.values():
+        for cid, call in self._calls.iteritems():
+            err = {'exception': 'RpcError',
+                   'message': 'Connection reset by peer'}
             if 'event' in call:
-                call['error'] = {'exception': 'RpcError',
-                                 'message': 'Connection reset by peer'}
+                call['error'] = err
                 call['return'] = None
                 call['event'].set()
+            else:
+                msg = {'id': cid, 'error': err, 'return': None}
+                self._manager.signal_arrival(msg)
         
         if callback is not None and not callable(callback):
             if self._handler is not None: