Move SharedMemoryLimits out of WebGraphicsContext3DCommandBufferImpl.

cc/output/delegating_renderer.cc

Index: cc/output/delegating_renderer.cc
diff --git a/cc/output/delegating_renderer.cc b/cc/output/delegating_renderer.cc
index a3746cc9d1b802cd36ff777d8e92bff009f1edb2..c063c7f608dfbfdbda048ebc7f942af291395b9b 100644
--- a/cc/output/delegating_renderer.cc
+++ b/cc/output/delegating_renderer.cc
@@ -48,22 +48,22 @@ DelegatingRenderer::DelegatingRenderer(RendererClient* client,
   if (!output_surface_->context_provider()) {
     capabilities_.using_shared_memory_resources = true;
   } else {
-    const ContextProvider::Capabilities& caps =
+    const auto& caps =
         output_surface_->context_provider()->ContextCapabilities();
 
-    DCHECK(!caps.gpu.iosurface || caps.gpu.texture_rectangle);
+    DCHECK(!caps.iosurface || caps.texture_rectangle);
 
-    capabilities_.using_egl_image = caps.gpu.egl_image_external;
-    capabilities_.using_image = caps.gpu.image;
+    capabilities_.using_egl_image = caps.egl_image_external;
+    capabilities_.using_image = caps.image;
 
     capabilities_.allow_rasterize_on_demand = false;
 
     // If MSAA is slow, we want this renderer to behave as though MSAA is not
     // available. Set samples to 0 to achieve this.
-    if (caps.gpu.msaa_is_slow)
+    if (caps.msaa_is_slow)
       capabilities_.max_msaa_samples = 0;
     else
-      capabilities_.max_msaa_samples = caps.gpu.max_samples;
+      capabilities_.max_msaa_samples = caps.max_samples;
   }
 }