request: allow fct_lump to lump factors based on order of appearance

[dan-reznik]

While fct_lump() currently allows us to keep the N most frequent factors (lumping the rest into "Other"), I would love to have an option to this function (call it "inorder") which would instead preserve the first N factors, lumping the rest into "Other".

library(magrittr)
library(forcats)
f <- c("hello","goodbye","ciao","arrivederci","ciao","hello") %>% fct_inorder
f
#> [1] hello       goodbye     ciao        arrivederci
#> [5] ciao        hello      
#> Levels: hello goodbye ciao arrivederci

To preserve the first two appearing levels of the factor, create function fct_lump_inorder(), all other levels are lumped into "Other":

f %>% fct_lump_inorder(2)
#> [1] hello       goodbye     Other       Other
#> [5] Other        hello      
Levels: hello goodbye Other

Notice this is very useful if I have reordered a factor using fct_reorder() and now only want the first N of the list (akin to arrange() + head()). Example:

df %>%
   mutate(col1 = col1 %>% fct_reorder(-col2) %>% fct_lump_inorder(4)) %>%
   ggplot(aes(...)) + ....