Increase performance when accessing underlying data

[TimTeaFan]

Currently most functions in the over-across function family access the underlying data with a call to dplyr::cur_data() which seems to be inefficient.

There exist other ways to access the underlying data which seem to be more performant. Tests are needed if they are in the same way reliable as a call to dplyr::cur_data()

Changing this should increase the performance of all functions that use dplyr::cur_data().