2711: do AABB removals by an unordered_map lookup rather than spatial search

[ericwa]

Fixes #2711

[ericwa]

I haven't tested this for memory leaks yet, and I guess the node heights aren't being tested.
Otherwise I like how this turned out. Now that the oldBounds aren't needed, it paves the way for a batch update API where AABBTree can decide internally whether to make incremental updates or rebuild if enough nodes are changing. Not that it's urgent or anything!

[kduske]

Looks good, very similar to what I had in mind, too. I just have a few minor comments.

[ericwa]

Thanks for the review. One last thing I wanted to do was forward-declare AABBTree and wrap it in a unique_ptr, now you can modify AABBTree and only 2 or 3 files recompile instead of dozens.