Bug 283286

For the planned threaded GPU rendering mode we need a way to mark a NativeImage/ImageBuffer "as done", to inject a fence in the GL command stream, that we can wait on at a later point in time to guarantee that all operations finished. Therefore I propose to add Skia specific functions to ImageBuffer(Backend)/NativeImageBackend: - finishAcceleratedRenderingAndCreateFence() Used to notify the underlying SkSurface, that it should flush all pending rendering operations and submit the work to the GPU. Afterwards inject a fence into the GL command stream, so that we we can wait upon completion at a later point, before attempting to use the ImageBuffer/NativeImage. - waitForAcceleratedRenderingFenceCompletion() Performs a 'serverWait()' operation on the GLFence, previously created by finishAcceleratedRenderingAndCreateFence(). This instructs to GPU to await completion of the previous rendering. finishAcceleratedRenderingAndCreateFence() will be used during DisplayList recording, whenever an ImageBuffer/NativeImage resource is used, upon entering the DisplayList::ResourceHeap cache. waitForAcceleratedRenderingFenceCompletion() will be used during DisplayList replaying, whenever an ImageBuffer/NativeImage resource is extracted from the DisplayList::ResourceHeap cache.