Per-point (not per-panel) expansion radii

@mattwala The approach of !95 (merged) gives us a truly local (i.e. per-point, not per-panel) quadrature resolution measure. We could use that to do per-point-varying expansion radii. What do you think?