Use the same buffer for arguments and return values

[JakeEhrlich]

Right now we use x0 to pass an array of arguments and we use x2 to pass the return buffer. The return buffer will (almost?) always be larger than the argument list because a guard could fail straight away. Additionally after the arguments are passed in we don't need to keep them. This would let us free up x2 and only x0 would be unavailable (which we could solve later by spilling x0)