Update and sync element unary/binary kernels

[elliottslaughter]

This is based on the assumption that ElementUnary is the template to follow (as it seems to be using the most modern code standards, as far as I can tell).

• Update the ElementBinary interface and implementation to match the style of ElementUnary
• Implement ElementBinary kernels for CPU (except for min/max)
• Add tests
• Minor updates to ElementUnary to make it more self-consistent (e.g., using consistent argument order everywhere)

I largely left the ElementBinary GPU kernel implementation alone; e.g., ElementUnary is a type-variadic implementation which I have made no effort to copy. However the GPU implementation should at least not regress and the external interface now matches the style of ElementUnary.

---

This change is