Consider deprecating convenience options from dfm()

[kbenoit]

Since we are on the issue of style, we could consider encouraging stepped dfm() processing options, versus using the dfm() convenience options. We would probably not want to remove or even deprecate (with warnings using .Deprecate()) them because so many people use them, but it seems better to use

dfm(toks) %>%
    dfm_remove(stopwords("en")) %>%
    dfm_group(by = "party")

than

dfm(toks, remove = stopwords("en"), groups = "party")

etc.

For educational materials and examples, at least, we could stick to the chained steps for clarity.

Just a thought, not essential.