HLSL: Don't emit the builtin PerVertex array as inout.

DXC barfs on this in certain cases. out seems to work. Passing this
around as value type is very questionable to begin with and there's no
way to express true references in HLSL :\

Author: HansKristian-Work

reference/opt/shaders-hlsl/mesh/mesh-shader-basic-lines.spv14.vk.nocompat.mesh

@@ -51,7 +51,7 @@ struct gl_MeshPerPrimitiveEXT
 
 groupshared float shared_float[16];
 
-void mesh_main(inout gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], inout gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint2 gl_PrimitiveLineIndicesEXT[22])
+void mesh_main(out gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], out gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint2 gl_PrimitiveLineIndicesEXT[22])
 {
     SetMeshOutputCounts(24u, 22u);
     float3 _173 = float3(gl_GlobalInvocationID);

reference/opt/shaders-hlsl/mesh/mesh-shader-basic-triangle.spv14.vk.nocompat.mesh

@@ -51,7 +51,7 @@ struct gl_MeshPerPrimitiveEXT
 
 groupshared float shared_float[16];
 
-void mesh_main(inout gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], inout gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint3 gl_PrimitiveTriangleIndicesEXT[22])
+void mesh_main(out gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], out gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint3 gl_PrimitiveTriangleIndicesEXT[22])
 {
     SetMeshOutputCounts(24u, 22u);
     float3 _29 = float3(gl_GlobalInvocationID);

reference/shaders-hlsl/mesh/mesh-shader-basic-lines.spv14.vk.nocompat.mesh

@@ -51,7 +51,7 @@ struct gl_MeshPerPrimitiveEXT
 
 groupshared float shared_float[16];
 
-void main3(inout uint2 gl_PrimitiveLineIndicesEXT[22], inout gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22])
+void main3(inout uint2 gl_PrimitiveLineIndicesEXT[22], out gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22])
 {
     gl_PrimitiveLineIndicesEXT[gl_LocalInvocationIndex] = uint2(0u, 1u) + gl_LocalInvocationIndex.xx;
     gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_PrimitiveID = int(gl_GlobalInvocationID.x);
@@ -61,7 +61,7 @@ void main3(inout uint2 gl_PrimitiveLineIndicesEXT[22], inout gl_MeshPerPrimitive
     gl_MeshPrimitivesEXT[gl_LocalInvocationIndex].gl_PrimitiveShadingRateEXT = int(gl_GlobalInvocationID.x) + 3;
 }
 
-void main2(inout gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], inout gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint2 gl_PrimitiveLineIndicesEXT[22])
+void main2(out gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], out gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint2 gl_PrimitiveLineIndicesEXT[22])
 {
     SetMeshOutputCounts(24u, 22u);
     gl_MeshVerticesEXT[gl_LocalInvocationIndex].gl_Position = float4(float3(gl_GlobalInvocationID), 1.0f);
@@ -81,7 +81,7 @@ void main2(inout gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], inout gl_MeshPerPri
     }
 }
 
-void mesh_main(inout gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], inout gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint2 gl_PrimitiveLineIndicesEXT[22])
+void mesh_main(out gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], out gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint2 gl_PrimitiveLineIndicesEXT[22])
 {
     main2(gl_MeshVerticesEXT, gl_MeshPrimitivesEXT, _payload, gl_PrimitiveLineIndicesEXT);
 }

reference/shaders-hlsl/mesh/mesh-shader-basic-triangle.spv14.vk.nocompat.mesh

@@ -51,7 +51,7 @@ struct gl_MeshPerPrimitiveEXT
 
 groupshared float shared_float[16];
 
-void mesh_main(inout gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], inout gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint3 gl_PrimitiveTriangleIndicesEXT[22])
+void mesh_main(out gl_MeshPerVertexEXT gl_MeshVerticesEXT[24], out gl_MeshPerPrimitiveEXT gl_MeshPrimitivesEXT[22], TaskPayload _payload, inout uint3 gl_PrimitiveTriangleIndicesEXT[22])
 {
     SetMeshOutputCounts(24u, 22u);
     gl_MeshVerticesEXT[gl_LocalInvocationIndex].gl_Position = float4(float3(gl_GlobalInvocationID), 1.0f);