Log HDFS Balancer exit message to its own log

Currently, the exit message is logged to stderr. It would be more convenient if we also log this to Balancer log when people want to figure out why Balancer is aborted.