`drop_na()` shouldn't drop empty atomic elements of list-cols

[DavisVaughan]

library(tidyr)

df <- tibble(x = list(1L, NULL, integer()))

drop_na(df)
#> # A tibble: 1 × 1
#>   x        
#>   <list>   
#> 1 <int [1]>

^{Created on 2021-11-19 by the reprex package (v2.0.1)}