[SPARK-2652] [PySpark] donot use KyroSerializer as default serializer

KyroSerializer can not serialize customized class without registered explicitly, use it as default serializer in PySpark will introduce some regression in MLlib.

cc mengxr

Author: Davies Liu <davies@databricks.com>

Closes #2916 from davies/revert and squashes the following commits:

43eb6d3 [Davies Liu] donot use KyroSerializer as default serializer

python/pyspark/context.py

@@ -43,7 +43,6 @@ __all__ = ['SparkContext']
 # These are special default configs for PySpark, they will overwrite
 # the default ones for Spark if they are not configured by user.
 DEFAULT_CONFIGS = {
-    "spark.serializer": "org.apache.spark.serializer.KryoSerializer",
     "spark.serializer.objectStreamReset": 100,
     "spark.rdd.compress": True,
 }