Refactor model pipeline in cesium

[danielzhong]

In the glTF edge visibility PR, there are cases where shader output declarations conflict with OIT’s output layout, even if the code paths are not executed in the OIT pass.

I introduced a temporary solution:
A new flag CESIUM_REDIRECTED_COLOR_OUTPUT was introduced to indicate that a shader has been derived/modified by Cesium’s rendering pipeline. This allows MRT features and other passes to conditionally declare output variables only when not conflicting with the redirected shader outputs.

However, we should consider a broader refactor of the model pipeline to handle this issue in a more systematic way.

#12859

[javagl]

Whenever there's an issue that starts with "Refactor...", my first reaction is: "Yes, that's most likely overdue". The main reason why there is so much need for refactoring is simple: Everything is connected to everything. The main reason why so few reactorings are happening is simple: Everything is connected to everything.

This issue is already assigned. This looks like it is supposed to be addressed soon. If it is supposed to be addressed, I'd be curious whether there already more specific plans or steps for this refactoring. (Maybe I'm misinterpreting that, and the first pass on this issue will exactly be to develop such a work plan...?)

[markschlosseratbentley]

@javagl: to try to address your comment:

@danielzhong, correct me if I am wrong, but I believe this issue could be considered to consist of two phases:

1. Figure out a better approach to handle the functionality that CESIUM_REDIRECTED_COLOR_OUTPUT achieves.
2. Implement that better approach.

[danielzhong]

@javagl @markschlosseratbentley You’re right, there isn’t a concrete plan yet, the first step will be to outline the refactor plan, and then we’ll move on to implementation.