Experiment with redefining bifunctors as uncurried functors

[Alizter]

In #1883 I mentioned that we should consider redefining Is0Bifunctor F as Is0Functor (ucurry F). This has a few advantages:

1. It allows us to use our "functorial-at-the-same-time" definitions.
2. We don't have extra terms with the fmap01 and fmap10 split.
3. Bifunctor coherence becomes a lemma and a way of constructing bifunctors.
4. It generalised much more easily to the ternary and n-ary case.

I think we would be able to make some arguments slicker by doing this, but I haven't spent anytime really checking. It may however present other problems that I haven't thought about. I'll record this issue so that we don't forget about it.