Fix 3D Tiles depth picking

[bagnell]

Fixes #5676, #5731, and #5683.

• Fixes picking depth on 3D Tiles.
• Adds classificationType property to GroundPrimitive and ClassificationPrimitive which specifies whether to classify terrain, 3D Tiles, or both.

[pjcozzi]

Looks OK to me at quick glance. Just one request: can you add - or update - a Sandcastle example that shows the differences between the different enum types, e.g., a polyline going over only terrain, only a tileset, or both.

@lilleyse can you review and verify each issue this fixed, and merge when ready? Thanks!

[lilleyse]

Code looks good and I confirmed the issues are fixed.

[lilleyse]

For consistency should this be renamed to CESIUM_3D_TILE to match Pass.CESIUM_3D_TILE and Pass.CESIUM_3D_TILE_CLASSIFICATION?

[bagnell]

@lilleyse This is ready for another look.

[lilleyse]

Could the rectangle in the demo be smaller so it doesn't cover the entire tileset? Maybe just the blue area:

[bagnell]

@lilleyse This is ready.

[bagnell]

@lilleyse Updated after our offline discussion.

[lilleyse]

Works well now.

@pjcozzi instead of offsetting the depth plane we now just render the depth plane after the 3D Tiles pass. A tileset on the opposite side of the globe would probably not be rendered at all due to SSE. Can you think of possible edge cases?

[lilleyse]

I'm going to merge, but if anyone has issues with the approach we can go back to offsetting the depth plane.

[pjcozzi]

@pjcozzi instead of offsetting the depth plane we now just render the depth plane after the 3D Tiles pass. A tileset on the opposite side of the globe would probably not be rendered at all due to SSE. Can you think of possible edge cases?

Doesn't this mean that most apps will need to enable depthTestAgainstTerrain?

[lilleyse]

They shouldn't need to. The main side effect with moving the depth plane after 3D Tiles classification was that tilesets might show through the opposite side of the globe when depthTestAgainstTerrain is false. But we figured that a tileset on the opposite side side would not get rendered at all because its root would either not be visible or would not meet SSE.

[pjcozzi]

Maybe this isn't a common case, but what if I have a label or a 3D model in drawn in the area of a 3D tileset, then I take a 45 degree horizon or top down view, wouldn't the depth values of the tileset be gone and the label/model would shown in front?

[lilleyse]

I'm pretty sure that case is fixed with #5789. Now the depth plane doesn't overwrite the 3D Tiles depth.

[pjcozzi]

OK, sounds good!