[SPARK-8141] [SQL] Precompute datatypes for partition columns and reuse it

JIRA: https://issues.apache.org/jira/browse/SPARK-8141

Author: Liang-Chi Hsieh <viirya@gmail.com>

Closes #6687 from viirya/reuse_partition_column_types and squashes the following commits:

dab0688 [Liang-Chi Hsieh] Reuse partitionColumnTypes.

sql/core/src/main/scala/org/apache/spark/sql/sources/interfaces.scala

@@ -435,8 +435,9 @@ abstract class HadoopFsRelation private[sql](maybePartitionSpec: Option[Partitio
           // partition values.
           userDefinedPartitionColumns.map { partitionSchema =>
             val spec = discoverPartitions()
+            val partitionColumnTypes = spec.partitionColumns.map(_.dataType)
             val castedPartitions = spec.partitions.map { case p @ Partition(values, path) =>
-              val literals = values.toSeq.zip(spec.partitionColumns.map(_.dataType)).map {
+              val literals = values.toSeq.zip(partitionColumnTypes).map {
                 case (value, dataType) => Literal.create(value, dataType)
               }
               val castedValues = partitionSchema.zip(literals).map { case (field, literal) =>