If a DataNode is down, then remove other DataNode may failed

In the current logic, if remove a DataNode X,  we will exec regions migrate on DataNode  Y or DataNode  Z,   (X, Y, Z has  same region replica) 

But before  move DataNode X, if DataNode Y or DataNode Z  is down,  maybe ConfigNode will  send regioin migrate task to the down node,  will cause remove DataNode X failed.

solution:  filter online DataNode to exec region migrate