Do not load prepass normals for transmissive materials (#11140)

Turns out whenever a normal prepass was active (which includes whenever
you use SSAO) we were attempting to read the normals from the prepass
for the specular transmissive material. Since transmissive materials
don't participate in the prepass (unlike opaque materials) we were
reading the normals from “behind” the mesh, producing really weird
visual results.

# Objective

- Fixes #11112.

## Solution

- We introduce a new `READS_VIEW_TRANSMISSION_TEXTURE` mesh pipeline
key;
- We set it whenever the material properties has the
`reads_view_transmission_texture` flag set; (i.e. the material is
transmissive)
- If this key is set we prevent the reading of normals from the prepass,
by not setting the `LOAD_PREPASS_NORMALS` shader def.

---

## Changelog

### Fixed

- Specular transmissive materials no longer attempt to erroneously load
prepass normals, and now work correctly even with the normal prepass
active (e.g. when using SSAO)

Author: coreh

crates/bevy_pbr/src/material.rs

@@ -606,6 +606,11 @@ pub fn queue_material_meshes<M: Material>(
             if mesh.morph_targets.is_some() {
                 mesh_key |= MeshPipelineKey::MORPH_TARGETS;
             }
+
+            if material.properties.reads_view_transmission_texture {
+                mesh_key |= MeshPipelineKey::READS_VIEW_TRANSMISSION_TEXTURE;
+            }
+
             mesh_key |= alpha_mode_pipeline_key(material.properties.alpha_mode);
 
             let pipeline_id = pipelines.specialize(

crates/bevy_pbr/src/render/mesh.rs

@@ -491,6 +491,7 @@ bitflags::bitflags! {
         const DEPTH_CLAMP_ORTHO                 = 1 << 10;
         const TEMPORAL_JITTER                   = 1 << 11;
         const MORPH_TARGETS                     = 1 << 12;
+        const READS_VIEW_TRANSMISSION_TEXTURE   = 1 << 13;
         const BLEND_RESERVED_BITS               = Self::BLEND_MASK_BITS << Self::BLEND_SHIFT_BITS; // ← Bitmask reserving bits for the blend state
         const BLEND_OPAQUE                      = 0 << Self::BLEND_SHIFT_BITS;                   // ← Values are just sequential within the mask, and can range from 0 to 3
         const BLEND_PREMULTIPLIED_ALPHA         = 1 << Self::BLEND_SHIFT_BITS;                   //
@@ -737,7 +738,7 @@ impl SpecializedMeshPipeline for MeshPipeline {
             // the current fragment value in the output and the depth is written to the
             // depth buffer
             depth_write_enabled = true;
-            is_opaque = true;
+            is_opaque = !key.contains(MeshPipelineKey::READS_VIEW_TRANSMISSION_TEXTURE);
         }
 
         if key.contains(MeshPipelineKey::NORMAL_PREPASS) {