Rewrite PMMLTransformer as a Scala class. Now supports Spark 2.x

[jmahonin]

• Adjust pom.xml to include scala compilation / test
• Use Spark's high-level udf() API instead of ScalaUDF
• Add unit test to check correctness