Instead of using a lambda in the order_by in csvkit, we could now delegate this to agate. However, csvkit's rule for None was that it is less than everything, but in agate, None is greater than everything. Would this be a breaking change, or can we change the rule?
/cc @onyxfish
Also, note that this PR may fail until agate is released with https://github.com/wireservice/agate/pull/607 merged.