[RFC] Should the GTK4 migration be reverted from master and put in it's own branch

I'll preface this with IN MY OPINION

The GTK4 migration is not going well. There are numerous regressions which lead me to question the amount of testing that is being done. I suspect there is a very heavy reliance on AI to generate correct code and little or no testing to make sure it's correct. I'm basing this on the number of problems discovered so far and how simple it was to discover them (basically just using darktable).

My biggest concern is that this isn't going to be completed/corrected by 5.8. My second concern is that it's affecting development of other features. I'm working on a solution to #21482 where I fire events for certain rating/color label/tagging/metadata/geolocation changes but I'm having trouble testing because I'm using the UI to trigger the event and discovering the UI doesn't work (click on a rating star or color label icon to assign it to a selected image).

The migration is also hampered by the fact we have no UI testing tools, so all the testing is user intensive and experience driven (the more you know the more you can test). I think I can automate some of the UI testing using Lua and darktable.gui.action() calls. What I can't emulate/test is mouse interactions. So, I could select an image, then activate the rating button to assign a star rating, then check the image to see if the rating was applied. What I can't check would be the mouse over the rating button and the click actually causing the rating button to perform an action.

Start by reading the GTK4 migration discussion and the testing concerns described for darktable's rating and color-label UI. Review the proposed Lua darktable.gui.action() approach and issue #21482; done means reaching and documenting a clear decision about reverting or branching the migration, with an agreed testing plan.