Add individual model projection in 2D / CV for `ModelExperimental

[j9liu]

This PR closes #9934. It allows models loaded in with ModelExperimental to render in 2D / CV, without needing projectTo2D enabled. This is ideal for individual models that aren't part of a 3D Tiles tileset.

This required refactoring the draw command system of ModelExperimental; now there is a ModelExperimentalDrawCommand class that handles a primitive's draw command, and any other commands derived from it.

I also found that the check for adding commands in 2D for Model was incorrect; it would always evaluate to true, so it was always adding the 2D commands even if the model was nowhere near the IDL.

Sandcastles for testing:

• 2D / CV
• toggle back face culling (to confirm that it works with new command system)
• toggle shadows (to confirm that it works with new command system)

@ptrgags can you review?

[cesium-concierge]

Thanks for the pull request @j9liu!

• ✔️ 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.

[ptrgags]

@j9liu looking at the sandcastles you posted, I had some questions:

2D / CV Sandcastle:

1. I notice that the lighting looks different on the two halves of the model in 2D mode (I presume because the light direction is different for each copy of the model?). Is this the intended behavior? I would think that you need both halves of the model should pretend the sun is in the same place, though not sure how easy it would be to override czm_lightDirection in the shader. Might be easiest to leave it alone, but wasn't sure.
   
2. I noticed some weird behavior in the map when I do the following in that sandcastle. Might just be a corner case in the camera code for 2D. Can you reproduce this behavior? What about with the old Model?

• Zoom out on the map so you can see the map repeat.
• Pan the view so you can see the IDL 180 degrees to the right of where you started
• Load a different model from the dropdown
• The map suddenly renders as 2 wedges for some reason:
  

3. I notice that the model always shows up on one edge of the map in CV mode (either the left or the right, not always consistently. e.g. the balloon shows up on the right but everything else shows up on the left. Is that what we want, or should it appear on both sides of the map?
4. In the culling/translucent Sandcastle, if enable both, the model shows up as translucent but back faces are not culled. Is that supposed to happen?

[ptrgags]

@j9liu looking at the sandcastles you posted, I had some questions:

2D / CV Sandcastle:

1. I notice that the lighting looks different on the two halves of the model in 2D mode (I presume because the light direction is different for each copy of the model?). Is this the intended behavior? I would think that you need both halves of the model should pretend the sun is in the same place, though not sure how easy it would be to override czm_lightDirection in the shader. Might be easiest to leave it alone.
   
2. I noticed some weird behavior in the map when I do the following in that sandcastle. Might just be a corner case in the camera code for 2D. Can you reproduce this behavior? What about with the old Model?

• Zoom out on the map so you can see the map repeat.
• Pan the view so you can see the IDL 180 degrees to the right of where you started
• Load a different model from the dropdown
• The map suddenly renders as 2 wedges for some reason:
  

3. I notice that the model always shows up on one edge of the map in CV mode (either the left or the right, not always consistently. e.g. the balloon shows up on the right but everything else shows up on the left. Is that what we want, or should it appear on both sides of the map?
4. In the culling/translucent Sandcastle, if enable both, the model shows up as translucent but back faces are not culled. Is that supposed to happen?

[j9liu]

@ptrgags

1. This behavior is consistent with Model. If we want, we can fix it later, but I don't think it should delay this PR.
2. Yeah I see this happen. I think it's because of the camera transformation when choosing a new model. The orientation of the camera is meant to be in 3D space but I see it causes problems for 2D mode. Probably should open an issue for this, but again outside the scope of the PR.
3. I believe this is consistent with Model.
4. This is supposed to happen. Back-face culling shouldn't apply to translucent models.

[ptrgags]

@j9liu I'm liking the new ModelExperimentalDrawCommand architecture, it feels a lot more straightforward than what Model was doing, and it simplifies quite a bit of the rest of the code.

I had some comments, mostly small suggestions or clarification questions

[ptrgags]

Also, is there a reason why this code isn't checking a similar conditions at y=0?

e.g. if left < 0 && right > 0 wouldn't that also be straddling the IDL?

[j9liu]

I'm not sure what you mean, does the IDL wrap around to y = 0?

[j9liu]

Okay, after some debugging, I found that it was supposed to check if left < -idl2D && right > -idl2D as well. so the general idea was right

[ptrgags]

@j9liu this is looking pretty close, though there's a small merge conflict in the changelog.

I also noticed that in 2D mode, the propellers on the drone disappear on one side:

I wondered if the individual primitives in the model are being translated to different copies of the IDL since each primitive bounding sphere is different. I tried modifying the Sandcastle to put the model close to the edge but not quite. I notice that at longitude = 179.999994 degrees this behavior appears, but when I change it to 179.999993 it works fine.

[j9liu]

@ptrgags updated!

[ptrgags]

@j9liu looks good and I confirmed that you fixed the missing primitives. Thanks!