Save 16 bytes per MeshUniform in uniform/storage buffers (#11999)

# Objective

- Save 16 bytes per MeshUniform in uniform/storage buffers.

## Solution

- Reorder members of MeshUniform to capitalise on alignment and size
rules for tighter data packing. Before the size of a MeshUniform was 160
bytes, and after it is 144 bytes, saving 16 bytes of unused padding for
alignment.

---

## Changelog

- Reduced the size of MeshUniform by 16 bytes.

crates/bevy_pbr/src/render/mesh.rs

@@ -201,6 +201,13 @@ pub struct MeshUniform {
     // Affine 4x3 matrices transposed to 3x4
     pub transform: [Vec4; 3],
     pub previous_transform: [Vec4; 3],
+    // 3x3 matrix packed in mat2x4 and f32 as:
+    //   [0].xyz, [1].x,
+    //   [1].yz, [2].xy
+    //   [2].z
+    pub inverse_transpose_model_a: [Vec4; 2],
+    pub inverse_transpose_model_b: f32,
+    pub flags: u32,
     // Four 16-bit unsigned normalized UV values packed into a `UVec2`:
     //
     //                         <--- MSB                   LSB --->
@@ -211,13 +218,6 @@ pub struct MeshUniform {
     //
     // (MSB: most significant bit; LSB: least significant bit.)
     pub lightmap_uv_rect: UVec2,
-    // 3x3 matrix packed in mat2x4 and f32 as:
-    //   [0].xyz, [1].x,
-    //   [1].yz, [2].xy
-    //   [2].z
-    pub inverse_transpose_model_a: [Vec4; 2],
-    pub inverse_transpose_model_b: f32,
-    pub flags: u32,
 }
 
 impl MeshUniform {

crates/bevy_pbr/src/render/mesh_types.wgsl

@@ -5,7 +5,6 @@ struct Mesh {
     // Use bevy_render::maths::affine_to_square to unpack
     model: mat3x4<f32>,
     previous_model: mat3x4<f32>,
-    lightmap_uv_rect: vec2<u32>,
     // 3x3 matrix packed in mat2x4 and f32 as:
     // [0].xyz, [1].x,
     // [1].yz, [2].xy
@@ -15,6 +14,7 @@ struct Mesh {
     inverse_transpose_model_b: f32,
     // 'flags' is a bit field indicating various options. u32 is 32 bits so we have up to 32 options.
     flags: u32,
+    lightmap_uv_rect: vec2<u32>,
 };
 
 #ifdef SKINNED