Allow functions and formulas with non-call RHS in across

[eutwt]

This PR does two things. Included together since they're both expanding possible across inputs and both pretty small changes.

1. Skips dt_squash_call for formulas whose RHS is not a call, allowing such formulas in across

• Behavior of non-call formulas in master shown here: mutate inconsistencies with dplyr #287

2. Allows anonymous non-formula functions in across

I could separate these if needed (and if there's interest in accepting either of them)

library(dplyr, warn.conflicts = FALSE)
library(dtplyr, warn.conflicts = FALSE)

lazy_dt(data.frame(a = 1, b = 2)) %>% 
  mutate(across(a, ~ b)) %>% 
  as_tibble()
#> # A tibble: 1 × 2
#>       a     b
#>   <dbl> <dbl>
#> 1     2     2

lazy_dt(data.frame(a = 1, b = 2)) %>% 
  mutate(across(a:b, function(x) x + 1)) %>% 
  as_tibble()
#> # A tibble: 1 × 2
#>       a     b
#>   <dbl> <dbl>
#> 1     2     3

^{Created on 2021-08-27 by the reprex package (v2.0.1)}

[hadley]

Looks good — just one question and can you please add a bullet to the top of NEWS.md? It should briefly describe the change and end with (@yourname, #issuenumber).

[hadley]

Could you construct the expected value a little more directly? I don't remember what quo_squash() does off the top of my head.

[eutwt]

Oops, it was a mistake to have it there, thanks for catching that. In this case it wasn't doing anything. I've removed it.

[eutwt]

Should be all set now