Closing a bypassed images starts recalculation again

[davidh]

It seems that recalculation of downstream images, after a bypassed image is closed, is back.
(Level, Crop, Color Balance, Brightness Curve...)

[jsachs]

Because images that depend on the image being closed via auxiliary inputs have their references reset rather than being closed, the propagation logic is now more complicated. Also, closing a single image is now just treated as a special case of closing a list of images, in which case the propagation logic is even more complicated. Consequently, this earlier optimization has been abandoned for now. It is one of a number of optimizations that I have on my list to implement at some future date.