Slice the start-points of the data, returning the specificed subset.

This returns a view of the data, not a copy. This means that as positions are calculated they will become available to other overlapping slices.