Update user API and op definition based on comments.

Author: liutongxuan

rfcs/20200411-fuse_recv.md

@@ -149,8 +149,7 @@ additional devices as much as possible.
 Example of how to enable the FuseRecv feature:
 
 ```
-  >>> config = tf.ConfigProto()
-  >>> config.graph_options.optimizer_options.experimental.do_fuse_recv = True
+  >>> tf.config.optimizer.set_experimental_options({"do_fuse_recv": True})
 ```
 
 ### Compatibility
@@ -164,17 +163,17 @@ Example of how to enable the FuseRecv feature:
 ## Detailed Design
 
 ### FuseRecv Op
-We introduce the FuseRecv Op and an RPC operation named FuseRecvTensorAsync in
-RemoteWorker and WorkerService. The FuseRecv Op definition is as follows:
+We introduce the _RecvV2 Op and an RPC operation named FuseRecvTensorAsync in
+RemoteWorker and WorkerService. The _RecvV2 Op definition is as follows:
 
 ```
-  >>> REGISTER_OP("FuseRecv")  
+  >>> REGISTER_OP("_RecvV2")
   >>>   .Output("tensor: tensor_type")
   >>>   .Attr("tensor_type: list(type)")
   >>>   .Attr("tensor_name: list(string)")
-  >>>   .Attr("send_device: list(string)")
-  >>>   .Attr("send_device_incarnation: list(int)")
-  >>>   .Attr("recv_device: list(string)")
+  >>>   .Attr("send_device: string")
+  >>>   .Attr("send_device_incarnation: int")
+  >>>   .Attr("recv_device: string")
   >>>   .Attr("client_terminated: bool = false")
   >>>   .SetIsStateful()
   >>>   .SetShapeFn(shape_inference::UnknownShape);