[TRACKING] Space Invasion in GIMP 3

This is a catch-all TODO-list for all the remaining points I want us to handle regarding color invasion in GIMP 3. Basically the goal is not to add new features but make sure that existing features produce good results and that various color-related features are color-generic:

• Color picking should pick in the input image color space (#7898 (closed))
• Color view widgets should store colors in the source color space (e.g. in case of colors picked from an image, the non-converted original color and color space should be stored) and displayed in the display space (!1137 (merged)).
• When relevant, color view widgets should be contextual (relatively to the active image), either by showing an out-of-gamut indication when relevant (while showing colors in display space even when out-of-gamut) or by converting colors to the active image's space (possibly truncating to gamut) (!1137 (merged)).
• Color selection in Colors dockable should have both generic space selection methods and selection relative to the active image.
• Palette should store input colors and space as-is too (cf. !843 (comment 1838225)).
• In-GIMP color palette creation should also be color-space aware.
• Make a decision regarding color space conversions at image loading (cf. #8494 (moved)). Probably doing nothing (i.e. keeping the color space as-is) should be the new default.
• Basic color management and soft-proofing should be verified (#8107 and #8269).
• Out-of-gamut on-canvas handling to be improved (#8381)
• All blend/composite modes algorithms should be verified for bugs (#11924).
• Review GUI for blend/composite modes to be more explicit while simpler to use; also we should check if it is possible for the Legacy modes to be converted to one of the new modes variant (#3901).
• Usage of GimpRGB everywhere should be replaced either by GeglColor or by a new GimpColor which should be generic and extensible. This should happen both in core and libgimp.
• The layer mode GUI must be updated to make blend/composite space/modes editing much more obvious and usable. (#3901)
• Blend Space perceptual seems broken with a non-perceptual imaged as what it does (didn't look at the code, just guessing) is actually to just work within the image space. So an "Addition" will just add layer values in whatever is the image space. In other words, if your image is stored linearly, it works the same in linear and perceptual blend space (obviously wrong) and worse, if you change your image encoding (e.g. convert to sRGB), the rendering will change! [pretty sure there was a report about this, but I can't find it anymore]

This list is meant to be extended, edited, amended as we go. Some parts of this list are already done in parts. Other non-listed items were also already implemented in the previous years. I have not added things which already work.