Boundaries for difference checking is wrong

[spillerrec]

When a movement value of for example 0.5 is chosen, GradientPlane only considers the dimension of one of the images which is fine if both input images are the same size, but this is especially not the case with AverageAligner. So the boundaries is likely not correct, which is very likely the root cause of #116.

So this needs to be fixed. But what exactly does a movement value of 50 % mean when the input images are of different sizes? Secondly there is a need to consider only a subarea for the actual movement. For the AverageAligner, the movement should be centered around the last image added. This should help with aligning much better.