Optimize color computation in prepare_uinodes (#7311)

# Objective
Speed up `prepare_uinodes`. The color `[f32; 4]` is being computed separately for every vertex in the UI, even though the color is the same for all 6 verticies.

## Solution
Avoid recomputing the color and cache it for all 6 verticies.

## Performance
On `many_buttons`, this shaved off 33% of the time in `prepare_uinodes` (7.67ms -> 5.09ms) on my local machine.
![image](https://user-images.githubusercontent.com/3137680/213862448-236ac6e4-040a-4c86-a801-b947d99cc581.png)

crates/bevy_ui/src/render/mod.rs

@@ -523,11 +523,12 @@ pub fn prepare_uinodes(
             uvs = [uvs[3], uvs[2], uvs[1], uvs[0]];
         }
 
+        let color = extracted_uinode.background_color.as_linear_rgba_f32();
         for i in QUAD_INDICES {
             ui_meta.vertices.push(UiVertex {
                 position: positions_clipped[i].into(),
                 uv: uvs[i].into(),
-                color: extracted_uinode.background_color.as_linear_rgba_f32(),
+                color,
             });
         }