update spark example sort-groupby example

[quasiben]

• use sparksession
• use cache instead of persist
• partition random DF
• add examples with dataframes to show serialization overhead