Add tokens_sample()

[koheiw]

We have dfm_sample() and corpus_sample(), but why not tokens_sample()?
I am tied of doing

toks <- tokens(data_corpus_big)
toks[sample(seq_along(toks), 10)]

just to check how tokenization is going.

[kbenoit]

Good idea. Let's add a

tokens_sample(x, size = ndoc(x), replace = FALSE, prob = NULL, by = NULL, ...)

that works just as corpus_sample() does to sample from documents.