Add translucent tile classification

[ebogo1]

• Adds TranslucentTileClassification and specs for handling translucent tile classification
• Adds DrawCommand. depthForTranslucentClassification and logic in Cesium3DTile for using it
• Adds handling of translucent classification in Scene and View

[cesium-concierge]

Thanks for the pull request @ebogo1!

• ✔️ Signed CLA found.
• ❔ CHANGES.md was not updated.
  • If this change updates the public API in any way, please add a bullet point to CHANGES.md.

Reviewers, don't forget to make sure that:

• Cesium Viewer works.
• Works in 2D/CV.
• Works (or fails gracefully) in IE11.

[lilleyse]

I pushed a premultiplied alpha fix similar to the one in #9398 (comment): 43f6efd. The colors should look much better now.

Before	After
Sandcastle

[lilleyse]

Overall this looks pretty good! At some point it would be nice to unify this with GlobeTranslucency since they do kind of the same thing, but I'm relatively happy with this approach for now.

I went through the checklist in #8726 and all of those situations worked great. There are a couple artifacts I noticed separately though:

1. The translucent classification framebuffer is always rendered above all translucent objects in the scene. I don't think this is fixable without some major changes to the approach. GlobeTranslucency doesn't have this problem because it renders classification to a framebuffer first and then creates a derived command for each TRANSLUCENT command that applies the classification. I.e. classification is not applied as a screen-wide post process.

Sandcastle

2. I see a slight color change when I pick at certain angles

@likangning93 @ebogo1 do you have any ideas about quick fixes for these? If there no obvious solution we should open a separate issue for each of these artifacts and move on.

[likangning93]

@ebogo1 here's a simpler localhost Sandcastle for stepping through the algorithm.

[ebogo1]

@lilleyse @likangning93 I reverted the changes to the noClassificationModels and classificationType flags that were once part of https://github.com/CesiumGS/cesium-analytics/pull/620/ (these will be part of #9399). Tested this in Sandcastle and it seemed to work fine, so I think this should be ready once CI passes on 0398f15.

[lilleyse]

I realized I broke some unit tests with the premultiplied alpha change earlier. I fixed those in 397ece9.

Thanks @ebogo1 and @likangning93!

[lilleyse]

@ebogo1 can you open two separate issues for #9399 (comment)?

[lilleyse]

The second issue in #9399 (comment) is actually pretty serious when picking on mouse move. I confirmed that this problem happens even before the premultiplied alpha fix.

Even though this PR is merged we should address this bug in a follow up PR. CC @likangning93 @ebogo1

Sandcastle

[lilleyse]

Just noticed another bug while testing #9398 (but not related to that PR)

In this Sandcastle if you repeatedly toggle tileset translucency on/off it will eventually crash.

[lilleyse]

After offline discussion, @ebogo1 can you do the following:

• Open issue for (1) in Add translucent tile classification #9399 (comment)
• Open issue for Add translucent tile classification #9399 (comment) (which is the same bug as (2) in the previous link). Label as priority - next release
• Open issue for Add translucent tile classification #9399 (comment). Label as priority - next release

[ebogo1]

Thanks @lilleyse! Opened the three issues above.

[midnight-dev]

The changelog entry for this PR has a minor typo.

Added support for for drawing ...

[lilleyse]

Thanks @midnight-dev, I pushed a fix directly to master: 0c31cdf