Optimization thought

Heap allocation is really expensive so I've been thinking that I could optimize the aabb tree by removing heap allocation or by creating the object in a reserved heap chunk (maybe an std::vector) and keeping track of what slots are free (maybe an std::queue<index>). Also there is a concept that come in mind and is the one of "Arena" I'm gonna take a look a that too since I think is what I am looking for. As the title says this is just a thought and reminder for later since I cant see any performance cost yet to do tests.