mat4_mul_mvp

[Minterl]

Addresses #124 I understand that this would just be another layer of abstraction and not an actual optimized algorithm, but the use case is common enough (given we do have functions for constructing rotation, perspective projection, etc. matrices) that it would make sense to have an official idiom for constructing an MVP matrix.

I am haven't tested this, but as for implementing an optimized, explicit three matrix multiplication algorithm, I would wager that the the inputs are generally small enough that something like parallel Strassen is generally unnecessary.