introduce particle_id_type instead of std::size_t about mjolnir HOT 1 CLOSED

I did some experiments and benchmarks that changes size_t to uint16_t in NeighborList and ExclusionList and found that it almost has no effect.

In the NeighborList, each neighbor has not only its index but also the parameter for the pair, such as sigma_i + sigma_j / 2 in Lennard-Jones potential. Because of the alignment, changing size_t to uint16_t does not reduce the size of each neighbor type at all. It may affect when we choose float instead of double, but I think the effect is small compared to the complexity and the efforts to maintain it.

In the ExclusionList case, it may affect because it contains indices directly, but the whole process does not take so much time.

It possibly accelerates simulations if there are many particles. I will keep the branch for the future. But currently, I will close this issue once.

from mjolnir.