in APRDF the limits upon smoothing are not meaningful

[kjappelbaum]

we conflate the correlated scale with the original length scale

[kjappelbaum]

rather switch to num_bins, as bin size in this correlated space does not make sense as it depends on the properties and aggregation method