Grape optimization

[MichalKononenko]

Here's how I understand it

The propagator for any unit time is piecewise continuous. That means that we can express it as a product of matrix exponentials. So instead of evaluating all the matrix exponentials individually, I summed the Hamiltonians and then did the matrix exponential.

Please break this. This is probably way too good to be true, but it works.

[MichalKononenko]

I just realized this really shouldn't be merged until we get a fuller optimization up and running.

[ihincks]

This is indeed too good to be true; it will break whenever the Hamiltonian does not commute with itself at different time steps. Generally, exp(A+B) != exp(A)exp(B) except when [A,B]=0.