Optimize Draw Occlusion algorithm

This patch is aimed to reduce the running time of draw occlusion algorithm.
In draw occlusion, quad++ is used to increase the iterator pointer to
visit the next draw quad. In reality, the quad++ calls
cc::ListContainer::Iterator::operator++ and creates a new Iterator which
points at the next draw quad. As a result, quad++ is always the top 3
expensive line in draw occlusion.

BEFORE:
91.61% viz::Display::RemoveOverdrawQuads(CompositorFrame)
  32.21% cc::ListContainer::viz::DrawQuad::Iterator
  29.26% cc::ListContainer::viz::SharedQuadState::end
  7.42% cc::ListContainerHelper::PositionInCharAllocator
  4.40% cc::ListContainer::viz::SharedQuadState::begin
  4.23% gfx::Transform(SkMatrix44)
  1.59% cc::ListContainer::viz::SharedQuadState::Iterator


AFTER:
91.51% viz::Display::RemoveOverdrawQuads(CompositorFrame)
 83.37% viz::Display::RemoveOverdrawQuads(CompositorFrame)
   32.04% cc::ListContainer::viz::SharedQuadState::end
   29.08% cc::ListContainer::viz::SharedQuadState::ConstIterator
   7.54% cc::ListContainerHelper::PositionInCharAllocator
   4.75% cc::ListContainer::viz::SharedQuadState::begin
   4.35% gfx::Transform(SkMatrix44)
   0.99% cc::ListContainer::viz::SharedQuadState::ReverseIterator

ref: https://docs.google.com/document/d/1gM4h1ca2jDWdXbYa9yjC53m0sJzaT9skqbPGnfDsdyI
Bug: 672929
Change-Id: I377ab3af3fdfa7e3d7ceb42dcfc2056de149424d
Reviewed-on: https://chromium-review.googlesource.com/782063
Commit-Queue: Yi Xu <yiyix@chromium.org>
Reviewed-by: danakj <danakj@chromium.org>
Cr-Commit-Position: refs/heads/master@{#518358}

components/viz/service/display/display.cc

@@ -576,7 +576,7 @@ void Display::RemoveOverdrawQuads(CompositorFrame* frame) {
       }
 
       if (!current_sqs_intersects_occlusion) {
-        quad++;
+        ++quad;
         continue;
       }
 
@@ -586,7 +586,7 @@ void Display::RemoveOverdrawQuads(CompositorFrame* frame) {
               transform, quad->visible_rect)))
         quad = pass->quad_list.EraseAndInvalidateAllPointers(quad);
       else
-        quad++;
+        ++quad;
     }
   }
 }