Fix #2207.

Probably not the best way, but I update summarise.sf() to allow for .by arguments.
The solution here is to store the row IDs of the original data in the .rows column by dplyr::summarise().

The disadvantage of this solution is that a column named .rows will be overwritten if the user has used it.
However, the column name .rows is not generally used,
since it cannot be used in tibble::tibble() and as a grouping variable for dplyr::group_by() .

I also noticed that summarise.sf() has an argument of .dots, probably past dplyr.
Now I have left it as is, but should I change the documentation as well?