[SPARK-2877] [SQL] MetastoreRelation should use SparkClassLoader when creating the tableDesc

JIRA: https://issues.apache.org/jira/browse/SPARK-2877

Author: Yin Huai <huai@cse.ohio-state.edu>

Closes #1806 from yhuai/SPARK-2877 and squashes the following commits:

4142bcb [Yin Huai] Use Spark's classloader.

sql/hive/src/main/scala/org/apache/spark/sql/hive/HiveMetastoreCatalog.scala

@@ -19,8 +19,6 @@ package org.apache.spark.sql.hive
 
 import scala.util.parsing.combinator.RegexParsers
 
-import org.apache.hadoop.fs.Path
-import org.apache.hadoop.hive.conf.HiveConf
 import org.apache.hadoop.hive.metastore.api.{FieldSchema, StorageDescriptor, SerDeInfo}
 import org.apache.hadoop.hive.metastore.api.{Table => TTable, Partition => TPartition}
 import org.apache.hadoop.hive.ql.metadata.{Hive, Partition, Table}
@@ -39,6 +37,7 @@ import org.apache.spark.sql.catalyst.rules._
 import org.apache.spark.sql.catalyst.types._
 import org.apache.spark.sql.columnar.InMemoryRelation
 import org.apache.spark.sql.hive.execution.HiveTableScan
+import org.apache.spark.util.Utils
 
 /* Implicit conversions */
 import scala.collection.JavaConversions._
@@ -288,7 +287,10 @@ private[hive] case class MetastoreRelation
   )
 
   val tableDesc = new TableDesc(
-    Class.forName(hiveQlTable.getSerializationLib).asInstanceOf[Class[Deserializer]],
+    Class.forName(
+      hiveQlTable.getSerializationLib,
+      true,
+      Utils.getContextOrSparkClassLoader).asInstanceOf[Class[Deserializer]],
     hiveQlTable.getInputFormatClass,
     // The class of table should be org.apache.hadoop.hive.ql.metadata.Table because
     // getOutputFormatClass will use HiveFileFormatUtils.getOutputFormatSubstitute to