Add column selector for `GroupBy.sortByKey`

[AndreiKingsley]

Now you can't specify key columns for sorting in sortByKey (all columns are used with the same order). No implementation is needed, need only to add an overload with column selector.

[Jolanrensen]

specifically, it needs a column selector for the GroupBy.keys columns. At the moment, this has the same type as the original dataframe, meaning it has accessors for columns that may not exist in the keys.

With the compiler plugin enabled, this is no issue.

Actually, come to think of it, you can already sort by specific key columns:

df.groupBy { a and b }.sortBy { a }

sortByKey() is simply a shortcut to sort for all key columns