Start Spark interpreter process using spark-submit

As Spark recommends use spark-submit to run spark application, Zeppelin can have an option to run spark interpreter process using spark-submit, after ZEPPELIN-160.

basic idea is, introduce environment variable 'USE_SPARK_SUBMIT'.

When SPARK_HOME and USE_SPARK_SUBMIT is defined, that interpreter.sh launches spark interpreter process using spark-submit.

This could help remove some problems oriented from Zeppelin starts SparkContext not exactly the same way spark-submit does.