Add min_freq and freq_type to textstat_collocations()

[koheiw]

It is sometimes difficult to set min_count because we don't know how many collocations in the corpus. If the number is too low, we have to wait long time to get the result of computation.

How about adding min_freq and freq_type = c("count", "prop", "rank", "quantile") in a similar way to dfm_trim()? It is only to set min_count besed on the distribution in counts_seq.

quanteda.textstats/src/collocations.cpp

Lines 287 to 290 in 68a8489

	for (auto it = counts_seq.begin(); it != counts_seq.end(); ++it) {
	// conver to a vector for faster itteration
	seqs_all.push_back(std::make_pair(it->first, it->second.first));
	if (it->second.first < count_min) continue;