Add support for 3DTILES_bounding_volume_cylinder

[j9liu]

Depends on #1104 so merge that first. But I'm going to mark this as a draft PR because I think there's a spec question to be had.

This PR adds support for the 3DTILES_bounding_volume_cylinder extension found here. The extension is mainly used to support cylinder voxels in the 3DTILES_content_voxels extension.

The cylinder is meant to be combined with implicit octree tiling, but that results in some weirdness for the BoundingCylinder implementation. When other volumes are subdivided (e.g. box, region) they are still the same shape. But the sub-regions of a cylinder are only parts of a cylinder, not fully cylinders itself. The 3DTILES_bounding_volume_cylinder extension does not currently indicate a way to divide that initial cylinder.

The thing to do here, regardless of spec, is probably to rename BoundingCylinder -> BoundingCylinderRegion, to more accurately indicate that the volume may be a subregion of the fully cylinder. And that way, the division of a BoundingCylinderRegion into sub-regions is logically correct. But I'm wondering if the 3DTILES_bounding_volume_cylinder should also allow angular / radial portions itself? Or, will we restrict it to being a full cylinder, but note that implementations should be wary of handling subregions with angular / radial bounds.

[j9liu]

I also should explicitly note that for efficiency / time reasons, the underlying implementation of BoundingCylinder relies on OrientedBoundingBox as an approximation.

I had actually added cylinder-specific math to the compute distance function (derived from this paper) in this (e5e323f). Unfortunately I had to remove it when I remembered that we're often dealing with cylinder subregions instead of the full cylinder. The math might carry over still but I haven't worked it out.

[j9liu]

@lilleyse and I are reworking the spec here: CesiumGS/3d-tiles#791

Once that's merged in, I'll make changes here accordingly

[j9liu]

I'm marking this as ready for review. The BoundingCylinder class is now BoundingCylinderRegion, with fields that mirror the ones in CesiumGS/3d-tiles#791. It should also interface more cleanly with ImplicitTilingUtilities.

[kring]

A few comments (mostly nitpicking include style), but overall looks great!

[j9liu]

Thanks for the thorough review @kring ! I think I addressed all your comments, let me know if anything else comes up.

[kring]

Thanks @j9liu!