Fix USERDATA built-ins for GLES3 particle shaders

[permelin]

When using any of USERDATA1-6 in a particle shader with the compatibility renderer, every particle except the first one would glitch out due to a buffer stride mismatch.

This PR addresses that and some related issues:

1. When setting flags in specialization for active userdatas, the wrong bit offset was used.
2. Userdata values were not carried over between frames.
3. Enabling individual userdatas was done only based on the count of userdatas in the particle shader. So if the shader only used for example USERDATA6 that would give a count of one and hence USERDATA1 was enabled instead, causing the shader to fail to compile.

NOTE: To make it possible to use any of the six userdatas without requiring that 1 is used before 2 and so on, I had to remove the layout locations on the inputs. I don't have nearly enough knowledge to understand if that is safe or if there is a better way. I need a maintainer to give feedback on this.

[clayjohn]

NOTE: To make it possible to use any of the six userdatas without requiring that 1 is used before 2 and so on, I had to remove the layout locations on the inputs. I don't have nearly enough knowledge to understand if that is safe or if there is a better way. I need a maintainer to give feedback on this.

I think this will be fine. The GL spec allows you to specify binding points in the shader, using glVertexAttrib, or by allowing the driver to figure it out. In this case, you are allowing the driver to figure it out. For a simple shader like this, I think that is a totally fine choice.

[clayjohn]

Looks great! Tested locally and it works.

[akien-mga]

Thanks!

[mikejohnstn]

Thanks for finding this. Even trying to set USERDATA1 (not even read it) was causing all particles after the first to glitch for me in compatibility. This fixed it.