Colorbalancergb: rework the mask method

[aurelienpierre]

1. Add a graph to show the mask opacities (y axis : 100% = fully opaque, 0% = fully transparent) for the generated shadows/midtones/highlights masks, as a function of luminance (x axis) :
   
2. Add image previews for the mask areas (example : shadows vs. highlights), showing the output of the module :
   
3. Add a middle-grey fulcrum parameter to slide the middle-grey reference (where both shadows and highlights masks have each a 50% opacity)
4. Rework the midtones masking for better separation and transitions.
5. Re-order and rename GUI controls for better understanding and accurate color-science references.

Example work produced with this new method :

[TurboGit]

Not tested yet, some comments.

[aurelienpierre]

@TurboGit using for_four_channels here doesn't work since the whole loop is SIMD (see CI failing).

[TurboGit]

@TurboGit using for_four_channels here doesn't work since the whole loop is SIMD (see CI failing).

Sorry, I didn't see that from the diff :( Will test soon.

[elstoc]

This is the one thing I never understood with the olde worlde color balance - where the shadows/mid-tones/highlights fell - nice to be able to see/adjust it now.

[aurelienpierre]

This is the one thing I never understood with the olde worlde color balance - where the shadows/mid-tones/highlights fell - nice to be able to see/adjust it now.

Don't get the wrong idea : the old color balance does not have explicit luminance masking, so it's simply not possible to present the splitting like that because there is no splitting at all. Old color balance does out = ((in + offset) * slope)^power, and that's it. Each parameter has more weight on some part of the range, but they are all applied on the whole range.

[TurboGit]

Manually merged. Thanks.