[SQL] Fix Performance Issue in data type casting

Using lazy val object instead of function in the class Cast, which improved the performance nearly by 2X in my local micro-benchmark.

Author: Cheng Hao <hao.cheng@intel.com>

Closes #679 from chenghao-intel/fix_type_casting and squashes the following commits:

71b0902 [Cheng Hao] using lazy val object instead of function for data type casting

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/Cast.scala

@@ -182,7 +182,7 @@ case class Cast(child: Expression, dataType: DataType) extends UnaryExpression {
     case x: NumericType => b => x.numeric.asInstanceOf[Numeric[Any]].toFloat(b)
   }
 
-  def cast: Any => Any = dataType match {
+  private lazy val cast: Any => Any = dataType match {
     case StringType => castToString
     case BinaryType => castToBinary
     case DecimalType => castToDecimal