Using squared distances vs. squared cutoff in force and energy loops

[g-bauer]

Currently, we check the distance between interaction sites versus the cutoff.

let r = system.cell().distance(posi, posj);
if r >= self.cutoff {
    0.0
} else {
// compute potential / force 

We could test if using the squared distance and squared cutoff (an hence skipping calculation of the square root) impacts performance.

let r2 = system.cell().distance2(posi, posj);
if r2 >= self.cutoff2 {
    0.0
} else {
// (compute distance)
// compute potential / force