`n` argument for `step_other`

[brshallo]

Would like if threshold argument within step_other() could be specified by an integer (rather than just by proportion), or had a separate n argument to specify the minimum sample size before it is collapsed into the "other" category.
E.g. to specify minimum sample size of 30:

recipe(price ~ clarity + color + carat, data = diamonds) %>% 
  step_other(all_nominal(), n = 30)

I'm curious what the best way of doing this is currently? (Link to hack for specifying consistent sample size to step_other across datasets of different sizes.)

[brshallo]

Using a step_mutate() with forcats::fct_lump_min() it seems would accomplish this. Though is there a way to apply step_mutate() over all_nominal(), in the way one might with mutate_if()/_at()/_all() ?

[topepo]

Instead of adding a new argument, we could just use threshold and, if it is >1, treat it as a frequency.