Minor Description
A lots of Flink applications run on Hadoop. Flink will invoke Hadoop caller context APIs to set up its caller contexts in HDFS/Yarn, so Hadoop should add Flink as one of the users in the LimitedPrivate scope.
Considering that there are more and more upstream programs（e.g. HBase, HDFS, Hive, MapReduce, Pig, Spark, Flink）that rely on CallerContext， Making it public+evolving makes sense.