Reimplement R2 and R3 as V2 and V3

[isovector]

This PR replaces R2 and R3 with V2 R and V3 R respectively. There are some major benefits from this:

• These are proper numeric types and don't require the terrifying orphan instances from ImplicitCad imports orphan num instances #292
• These are proper functor types, meaning we can eliminate a good chunk of helper functions --- things like both, allthree, CompontwiseMultiplable, and in a follow-up PR, VectorStuff.
• These are strict types, meaning they are much better at preventing the usual Haskell memory leaks.

In addition, these vector types are well supported by the linear library, which in the future should allow us to simplify a good chunk of the math in this library and delegate it to linear.

NOTE I expected this to be a no-op change, but it has caused three of the golden tests to fail. They are visible indistinguishable, but it's still a little disconcerting. I've looked through the diff three times and can't spot the culprit.

Fixes #286
Fixes #292

[julialongtin]

ok, this is 'major release bump' territory. it completely blows up hslice, too. :)

[isovector]

I realized this morning we can add a typeclass class IsVector from to for converting from the old tuple representations to the new vector ones. That will prevent the public interface from changing at least.

How does this blow up hslice?