Currently, the logic to decode JPEGs using the VAAPI is mixed with a lot of VideoFrame-dependent code.

In order to implement gpu::ImageDecodeAcceleratorWorker to do general JPEG decoding, it would be better to extract VideoFrame-independent code into a common base class that can be used by both VaapiJpegDecodeAccelerator and the eventual implementation of gpu::ImageDecodeAcceleratorWorker (which doesn't need VideoFrames).