Fix bug in SmallStep method for particle advection

When a particle approaches the boundary of data set, Integrator::SmallStep() uses a binary search to find the largest time step that is within the dataset. Then, it uses the upper bound of the binary search to take an Euler step that pushes the particle "just outside" of the data block. There was a bug in the particle position that was used to take this step. This MR fixes this bug.

This MR got into a weird state after I tried to fix a commit message that was too short. I created a new MR for this. new MR/commit. And so will close this one.