2d Optical-flow stabalization about gyroflow

[AdrianEddy]

See #45

[siefkenj]

@AdrianEddy Perhaps you missed the 2d label on this issue? Currently when doing optical-flow stabilization Gyroflow seems to rotate in 3 dimensions to try and stabilize the video. I am hoping for stabilization done by only x/y translation and roll adjustments.

[AdrianEddy]

the 2d part is the reason why #45 is still open. Quote from my comment there: Proper implementation should focus more on better optical flow tracking, and there should be a second optical flow stabilization on top of it (using motion vectors to move small pixel groups, like Warp Stabilizer, instead relying only on estimated camera motion).