CompareImage: Feature a consistent autoscale

[vallsv]

Feature a consistent autoscale for the CompareImage.

Now

• the autoscale statistics are computer globally for the 2 images
• it is also computed on the raw image (instead of in the sliced left/right image)
• This makes the autoscale usable, from the ColormapDialog, including 3stddev
• subsequently when you are about to slice, the statistics are not recomputed -> that's faster

Changelog:

• CompareImage: Colormap autoscale is now global to the 2 raw images to compare
• CompareImage: updateColormap= was dropped, this can be set with getColormap()

[t20100]

There's a miss-spelling of pinned and probably a missing change regarding colormap2.

Otherwise, it's OK for me. We can live with ColormapWithPinnedData as long as it is not exposed publicly.

[vallsv]

This have to be rebaed on top of #3845

[vallsv]

I have committed an alternative implementation to drop the custom colormap class.

• The virtual item is used to hold the concatenation of the 2 images
• When the colormap is changed, an updated colormap is sealed with the range and set to the displayed images

That maybe a bit less efficient than the previous implementation, but this could be improved, and for now who cares.

The good thing is that the displayed histogram is properly computed from the concatenation (previously the histogram was computed from the crop of one of the images)

[t20100]

Can you rebase to include commit ccdc13b from main to fix CI on Windows?

[t20100]

Great! Colormap range handling is as expected, thanks!

I'd keep backward compatibility for the updateColormap arguments, it's doesn't hurt.

[vallsv]

While i was here, i have used the auto scale to improve the contrast on the composite image comparison.

The result is probably not exactly the same for the B&W composition. @payno would you have some times to check if it's fine for you? Else i anyway have to fix something because the B&W was not working anyway.

The good thing with this change is the colormap is used in a consistent way anyway the visualisation mode.

[vallsv]

@payno thinking about that a bit more, i think it is better to revert my change on the compare image for the B&W composition, it would be better to talk about that together before merging.

[t20100]

I'd be +1 to merge this PR as it was and work on additional feature/fix in another PR.

[vallsv]

@t20100 i am fine with it, let's merge it as it is

[t20100]

LGTM