Add context termination for History server on Yarn

The history server on Yarn only shows completed jobs. This adds a note concerning the needed explicit context termination at the end of a spark job which is a best practice anyway.
Related to SPARK-2972 and SPARK-3458

Author: moutai

docs/monitoring.md

@@ -150,6 +150,8 @@ follows:
 Note that in all of these UIs, the tables are sortable by clicking their headers,
 making it easy to identify slow tasks, data skew, etc.
 
+In Yarn, note that the history server only displays completed Spark jobs. One way to signal the completion of a Spark job is to stop the Spark Context explicitly (`sc.stop()`), or in Python using the `with SparkContext() as sc:` to handle the Spark Context setup and tear down, and still show the job history on the UI.
+
 # Metrics
 
 Spark has a configurable metrics system based on the