default clustering on large samples

[Duncan Ralph]

I recently changed the default behavior such that when clustering large samples (>10k seqs) --fast (i.e. vsearch clustering) is turned on by default, hopefully that doesn't cause too many issues for anybody. You can disable this with --no-naive-vsearch. The change is also reflected in the docs.