Look deeply into adding combine()-support in Spark MapReduce.

I have comments in the code that tell me that combine() support is not necessary. However, I don't really think this is true. Map processing seems to be taking longer than it should.