[SPARK-6483][SQL]Improve ScalaUdf called performance.

As issue [SPARK-6483](https://issues.apache.org/jira/browse/SPARK-6483) description, ScalaUdf is low performance because of calling *asInstanceOf* to convert per record.
With this, the performance of ScalaUdf is the same as other case.
thank lianhuiwang for telling me how to resolve this problem.

Author: zzcclp <[email protected]>

Closes #5154 from zzcclp/SPARK-6483 and squashes the following commits:

5ac6e09 [zzcclp] Add a newline at the end of source file
cc6868e [zzcclp] Fix for fail on unit test.
0a8cdc3 [zzcclp] indention issue
b73836a [zzcclp] Access Seq[Expression] element by :: operator, and update the code gen script.
7763848 [zzcclp] rebase from master

Author: zzcclp