Add additional compile time type safety to math instructions

[TheDan64]

This is a follow up to #45

We should constrain FloatMathValue & IntMathValue (names pending) to be implemented only for Vectors that have a numeric subtype, that is, IntValue<IntSize> and VectorValue<IntValue<IntSize>, VectorSize> for example.

If IntSize & VectorSize are not yet possible due to requiring const generics for example, we should split that off into a new issue for v0.3.0 and just implement for IntValue and VectorValue<IntValue> (and likewise for floats). See discussion in #45 for more info