clean up the logic of removing subregions with makeup/extra overlapping

`blabpy.seedlings.listened_time._remove_subregions` is unnecessarily complicated. The plan was to have multiple kinds of conditions: remove a subregion when it strictly partially overlaps with one of the, say, surpluses, contain a surplus that is fully nested, is fully nested within a surplus region, etc. The algorithm has been considerably simplified since then, so we only really need _overlaps_with_interval and it is only ever used once on a single set of `other_region_types` (makeup/surplus). This can all be simplified to something like `_remove_subregions_overlapping_with_makeup_or_surplus`.

If you get to this, it might be a good idea to break the process into logical, not technical steps: what we are
trying to do is figure out which subregions have been listened to: these are the ones with annotations or a special
comment except for those that only have annotations because they were used for makeup, which might have later been
renamed as surplus. So, it would make sense to have a `_remove_subregions_not_listened_to` function that would
account for annotations, special comments *and* overlapping makeup/surplus regions.

Another thing to consider: counting annotations after deoverlapping all the regions would be even better:
- annotations in makeup/surplus would not be in the subregions at all (overlaps are removed from subregions during
deoverlapping) so we would get the subregions without their own annotations removed for free,
- if those subregions have annotations outside of the makeup/surplus regions, we would probably want to know: in such
cases, the subregions would not get removed because they would still contain annotations ever after deoverlapping.