Imagine a case where two or more entities share the same vertex and index buffers and attributes, with the only difference being the byteOffset in the index attribute.

Here calculating a bounding volume based solely on the position attribute is wrong since it gives the exact same results for all the entities. Unfortunately this is exactly what calculateLocalBoundingVolume() does.

Ideally everything starting from BoundingVolumeCalculator should take the index attribute into account and, whenever the index attribute is present, vertex traversal should be done accordingly.