Support particle death

[urogdiaron]

Maybe swap dead particles to the end so we have contiguous alive ones.

When do we do that? Do we need to do that?

Right now the cpu side doesn't know how many particles are alive so it runs the compute shader for all possible particles either way. Maybe we should let the cpu side know first so such an optimization could make sense in the first place.

If the cpu side knew how many groups it needs to dispatch we would need the particle buffer to have the alive particles in the beginning.