Fix explicit resource allocator bug

This was folded into https://skia-review.googlesource.com/c/skia/+/142102 (Reduce arbitrary opList splitting when sorting (take 2)) but need not wait for that CL to land.

Change-Id: I883497129404fbf879e3c21f465f6debc5b5c2ed
Reviewed-on: https://skia-review.googlesource.com/143706
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
diff --git a/src/gpu/GrRenderTargetOpList.h b/src/gpu/GrRenderTargetOpList.h
index 8629a7c..81bd6f1 100644
--- a/src/gpu/GrRenderTargetOpList.h
+++ b/src/gpu/GrRenderTargetOpList.h
@@ -87,6 +87,9 @@
 
         op->visitProxies(addDependency);
         clip.visitProxies(addDependency);
+        if (dstProxy.proxy()) {
+            addDependency(dstProxy.proxy());
+        }
 
         return this->recordOp(std::move(op), caps, clip.doesClip() ? &clip : nullptr, &dstProxy);
     }