Convert var_name column to character before replacing NAs

[DavisVaughan]

We are updating tidyr::replace_na() to utilize vctrs, and that results in slightly stricter / more correct type conversions. See tidyverse/tidyr#1219

We noticed in revdeps that this package broke. An easy way to see this is by installing the PR mentioned above and running:

library(dplyr)
library(tidytidbits)

mtcars %>% count_at(vars(gear, cyl))
#> Warning: The `add` argument of `group_by()` is deprecated as of dplyr 1.0.0.
#> Please use the `.add` argument instead.
#> This warning is displayed once every 8 hours.
#> Call `lifecycle::last_lifecycle_warnings()` to see where this warning was generated.
#> Error: Can't convert `replace$gear` <character> to match type of `data$gear` <double>.

The problem boils down to the fact that you are calling replace_na() on numeric columns, but your replacement is a character string ("missing"). This is no longer allowed. It seems like you wanted the end result to be a character vector anyways, so this PR preemptively changes the !!var_name column to a character before applying replace_na().

We would greatly appreciate if you could merge this PR and submit a patch release of your package to CRAN so we can send tidyr in!