feature/add approx percentile

[MrPowers]

@nvander1 - looks great!

Couple of discussion points:

• can we call this function bebe_approx_percentile? I'd like to let users run import org.apache.spark.sql.BebeFunctions._ without getting confused about where these functions are coming from

• Can we use default parameter values rather than overloading? I'm not sure why the Scala style guide says to avoid default arguments. Probably because Java doesn't support default arguments and the method overloading is the only way to provide them with a sensible API.

Let me know your thoughts!

[MrPowers]

@nvander1 - merged, great contribution, thank you!!!