remove_parents in feature detection only considers overlapping features detected at the previous threshold

[w-k-jones]

The remove_parents function only compares features with those at the previous threshold value. The assumption that any feature detected at a certain threshold must also be true with a fixed n_min_threshold across all thresholds, but can be false if a range of n_min_threshold is provided (see e.g. the final notebook example in #283)

We should modify the way we store threshold regions in feature detection so that all previously detected features are considered by remove_parents, not just those at the previous threshold.

[freemansw1]

Thanks for this report; I don't think we considered this downstream consequence of #283 and the change to n_min_threshold.

We should probably hold off merging #283 until this is resolved.

[freemansw1]

Given that this has been resolved in #283, I'm going to close this issue.