Reduce the columns/fields of shuffled record

1. Group By without removing unused fields

On the flow above, we see that after Group By, only fields category_id and likes,
but all fields are shuffled. So it takes resource, and reduce performance.

2. Group By with removing unused field

On the flow above, before Group By, all the fields are removed except fields category_id and likes.
So the number of shuffled fields are reduced. It will increase performance.