Fixed _WorldToShadow warning.

com.unity.render-pipelines.lightweight/LWRP/Passes/DirectionalShadowsPass.cs

             float invHalfShadowAtlasWidth = 0.5f * invShadowAtlasWidth;
             float invHalfShadowAtlasHeight = 0.5f * invShadowAtlasHeight;
             cmd.SetGlobalTexture(destination.id, m_DirectionalShadowmapTexture);
-            if (shadowData.directionalLightCascadeCount > 1)
-                cmd.SetGlobalMatrixArray(DirectionalShadowConstantBuffer._WorldToShadow, m_DirectionalShadowMatrices);
-            else
-                cmd.SetGlobalMatrix(DirectionalShadowConstantBuffer._WorldToShadow, m_DirectionalShadowMatrices[0]);
+            cmd.SetGlobalMatrixArray(DirectionalShadowConstantBuffer._WorldToShadow, m_DirectionalShadowMatrices);
             cmd.SetGlobalVector(DirectionalShadowConstantBuffer._ShadowData, new Vector4(light.shadowStrength, 0.0f, 0.0f, 0.0f));
             cmd.SetGlobalVector(DirectionalShadowConstantBuffer._DirShadowSplitSpheres0, m_CascadeSplitDistances[0]);
             cmd.SetGlobalVector(DirectionalShadowConstantBuffer._DirShadowSplitSpheres1, m_CascadeSplitDistances[1]);

com.unity.render-pipelines.lightweight/LWRP/ShaderLibrary/Shadows.hlsl

 // Last cascade is initialized with a no-op matrix. It always transforms
 // shadow coord to half(0, 0, NEAR_PLANE). We use this trick to avoid
 // branching since ComputeCascadeIndex can return cascade index = MAX_SHADOW_CASCADES
-#ifdef _SHADOWS_CASCADE
-#else
-float4x4    _WorldToShadow;
-#endif
 float4      _DirShadowSplitSpheres0;
 float4      _DirShadowSplitSpheres1;
 float4      _DirShadowSplitSpheres2;
 
 half ComputeCascadeIndex(float3 positionWS)
 {
-    // TODO: profile if there's a performance improvement if we avoid indexing here
     float3 fromCenter0 = positionWS - _DirShadowSplitSpheres0.xyz;
     float3 fromCenter1 = positionWS - _DirShadowSplitSpheres1.xyz;
     float3 fromCenter2 = positionWS - _DirShadowSplitSpheres2.xyz;
     half cascadeIndex = ComputeCascadeIndex(positionWS);
     return mul(_WorldToShadow[cascadeIndex], float4(positionWS, 1.0));
 #else
-    return mul(_WorldToShadow, float4(positionWS, 1.0));
+    return mul(_WorldToShadow[0], float4(positionWS, 1.0));
 #endif
 }