ZEPPELIN-3296. Reorg livy integration test to minimize livy session

.travis.yml

@@ -114,17 +114,16 @@ matrix:
       dist: trusty
       env: PYTHON="3" SCALA_VER="2.10" SPARK_VER="1.6.3" HADOOP_VER="2.6" PROFILE="-Pspark-1.6 -Phadoop2 -Phadoop-2.6 -Pscala-2.10" SPARKR="true" BUILD_FLAG="package -DskipTests -DskipRat -am" TEST_FLAG="test -DskipRat -am" MODULES="-pl spark/interpreter,spark/spark-dependencies" TEST_PROJECTS="-Dtest=org.apache.zeppelin.spark.* -DfailIfNoTests=false"
 
-    # Test python/pyspark with python 2, livy 0.5
     - sudo: required
       dist: trusty
       jdk: "openjdk7"
-      env: PYTHON="2" SCALA_VER="2.10" SPARK_VER="1.6.3" HADOOP_VER="2.6" LIVY_VER="0.5.0-incubating" PROFILE="-Pspark-1.6 -Phadoop2 -Phadoop-2.6 -Pscala-2.10" BUILD_FLAG="install -am -DskipTests -DskipRat" TEST_FLAG="verify -DskipRat" MODULES="-pl livy" TEST_PROJECTS="-Dpyspark.test.exclude='' -DfailIfNoTests=false"
+      env: PYTHON="2" SPARK_VER="1.6.3" HADOOP_VER="2.6" LIVY_VER="0.5.0-incubating" PROFILE="" BUILD_FLAG="install -am -DskipTests -DskipRat" TEST_FLAG="verify -DskipRat" MODULES="-pl livy" TEST_PROJECTS=""
 
-    # Test python/pyspark with python 3, livy 0.5
+    # Test livy 0.5 with spark 2.2.0 under python3
     - sudo: required
       dist: trusty
-      jdk: "openjdk7"
-      env: PYTHON="3" SCALA_VER="2.11" SPARK_VER="2.0.0" HADOOP_VER="2.6" LIVY_VER="0.5.0-incubating" PROFILE="-Pspark-2.0 -Phadoop3 -Phadoop-2.6 -Pscala-2.11" BUILD_FLAG="install -am -DskipTests -DskipRat" TEST_FLAG="verify -DskipRat" MODULES="-pl livy" TEST_PROJECTS="-Dpyspark.test.exclude='' -DfailIfNoTests=false"
+      jdk: "openjdk8"
+      env: PYTHON="3" SPARK_VER="2.2.0" HADOOP_VER="2.6" LIVY_VER="0.5.0-incubating" PROFILE="" BUILD_FLAG="install -am -DskipTests -DskipRat" TEST_FLAG="verify -DskipRat" MODULES="-pl livy" TEST_PROJECTS=""
 
 before_install:
   # check files included in commit range, clear bower_components if a bower.json file has changed.