Fix pick with skyatmosphere

[rutsky]

Add raising DeveloperError if previously picked object is picked again in drillPick.
Don't draw SkyAtmosphere when picking.

Fixes #4783 and #4784.

[rutsky]

Please review changes thoroughly --- I'm not familiar with CesiumJS internals.

[rutsky]

What is expected pickId when picking not-object (stars, Earth tiles)? Shouldn't Context.getObjectByPickColor get zero color for anything but not object?

Because I see that Context.getObjectByPickColor gets colors like the following:

4278190338
4278190081
4278255873
4278190081
4278190338

[hpinkos]

Thanks @rutsky! Most of the team is out of the office this week, but we'll try review this soon.

[pjcozzi]

This would mask a bug elsewhere - basically setting a primitive's show to false not working. Can you remove this and keep the scope of this bug to picking the atmosphere?

[rutsky]

I excluded commit with adding DeveloperError from the pull request.

Are you sure there is no benefit in adding such internal consistency check? (I can open second PR with this change.)
IMO this check costs almost nothing, but can improve debugging experience next time when pick issue will arise.

[pjcozzi]

@rutsky usually we would add a new test, but it will be hard to reasonably do so without tightly coupling the test to the engine code so don't worry about it.

[pjcozzi]

What is expected pickId when picking not-object (stars, Earth tiles)?

Looking at the sky box, sun, and moon update functions they do not generate pick commands so they should not be pickable (they are handled outside our normal command processing so the color commands are still being executed in the pick pass, which is the problem).

The globe tiles won't generate any pick commands, other than writing depth, but they are processed in the right place so they are unlikely to be an issue.

I can finish this PR up next week if that is OK with you. Just address #4790 (comment) in the meantime, please.

[rutsky]

I can finish this PR up next week if that is OK with you. Just address #4790 (comment) in the meantime, please.

Thanks, @pjcozzi , it's better if you finish this PR, since it requires deeper knowledge of Cesium internals that I have, plus it's holidays here in Russia next week.

I removed commit with consistency check --- see my comment.

[pjcozzi]

This comment, #4783 (comment), was meant for this PR:

Thanks again for the fix, @rutsky. I made a small tweak to also not render the sun, moon, and sky box and merged everything in #4806 since I could not commit to your branch (could have been an issue on my end).