[SPARK-22965] [PySpark] [SQL] Add deterministic parameter to registerJavaFunction

[gatorsmile]

What changes were proposed in this pull request?

To register a JAVA UDF in PySpark, users are unable to specify the registered UDF is not deterministic. The proposal is to add the extra parameter deterministic at the end of the function registerJavaFunction

Below is an example.

>>> from pyspark.sql.types import DoubleType
>>> sqlContext.registerJavaFunction("javaRand",
...   "test.org.apache.spark.sql.JavaRandUDF", DoubleType(), deterministic=False)
>>> sqlContext.sql("SELECT javaRand(3)").collect()
[Row(UDF:javaRand(3)=3.12345)]

How was this patch tested?

Added test cases

[gatorsmile]

Although this is private and only used for PySpark, the JAVA users still could call it. For not breaking the existing behavior, I keep the original one unchanged.

[SparkQA]

Test build #85706 has finished for PR 20162 at commit cf0076b.

• This patch fails Python style tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #85707 has finished for PR 20162 at commit d63e816.

• This patch fails Python style tests.
• This patch merges cleanly.
• This patch adds no public classes.

[SparkQA]

Test build #85708 has finished for PR 20162 at commit 7e4f3c0.

• This patch fails PySpark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.