Compare speed to rsparse implementation

[alexpghayes]

I'm curious how much slower we are and what we can do about it. Not to compete with them but just because simulations using fastadi right now are somewhat painful.