arrow

ARROW-8758: [R] Updates for compatibility with dplyr 1.0

I tested this locally with the current version of `dplyr` on CRAN and the dev version scheduled to be released to CRAN on May 15. Our tests now pass with both versions.

Changes addressed:

* `group_by` now requires a character vector of grouping variable names, so now we use `group_vars()` instead of `groups()`. `group_vars()` works in the current `dplyr` release, so this is a simple change.
* The argument name in `group_by()` changed from `add` to `.add`, and calling it with the name that works in the current version raises a deprecation warning in dplyr 1.0. The fix here supports both spellings of the argument, and it avoids the warning by determining which version of the internal dplyr function exists and calling the appropriate one.
* `dplyr::transmute()` no longer calls `dplyr::mutate()` internally, so it doesn't just work on Arrow objects anymore. I skipped the one test that called it and left a TODO to add a transmute method.

Closes #7147 from nealrichardson/dplyr-1.0

Authored-by: Neal Richardson <neal.p.richardson@gmail.com>
Signed-off-by: Neal Richardson <neal.p.richardson@gmail.com>