Deprecate `cross()` and `cross_df()`

[lionel-]

• They are super slow: cross_df() is extremely slow compared to expand.grid() and crossing() #406
• They have bugs: error with cross_d when a list of same-type vectors is in the .l argument #327, cross_df doesnt preserve ordered factor type. #760

There is tidyr::expand_grid() which doesn't have these issues. The ordering of combinations is not the same, but probably not a big deal:

data <- list(
  id = c("John", "Jane"),
  greeting = c("Hello.", "Bonjour."),
  sep = c("! ", "... ")
)

data %>%
  cross() %>%
  map_chr(lift(paste))
#> [1] "John! Hello."     "Jane! Hello."     "John! Bonjour."   "Jane! Bonjour."
#> [5] "John... Hello."   "Jane... Hello."   "John... Bonjour." "Jane... Bonjour."

tidyr::expand_grid(!!!data) %>%
  pmap_chr(paste)
#> [1] "John! Hello."     "John... Hello."   "John! Bonjour."   "John... Bonjour."
#> [5] "Jane! Hello."     "Jane... Hello."   "Jane! Bonjour."   "Jane... Bonjour."