Implement click-move-click & snapping behavior for the georeferencer's move GCP point tool

[nirvn]

Description

This PR upgrades the georeferencer's move GCP point tool to rely on a click-move-click behavior. On top of harmonizing the behavior with other QGIS tools (such as the vertex editor tool, annotation node manipulation tool, etc.), it allows us to implement snapping when moving the GCP points.

Having moved to the click-move-click behavior also made it possible to implement a right-click cancel functionality when users realize they are moving the wrong GCP point.

Finally, GCP points now react to being hovered, with a larger symbol being drawn when that's the case. It makes for a much nicer experience trying to pick the right point to move.

[github-actions]

🪟 Windows builds

Download Windows builds of this PR for testing.
Debug symbols for this build are available here.
(Built from commit d6cf21b)

🪟 Windows Qt6 builds

Download Windows Qt6 builds of this PR for testing.
(Built from commit d6cf21b)

[nirvn]

@uclaros , was this PR ok to merge otherwise? :)

[uclaros]

@uclaros , was this PR ok to merge otherwise? :)

I now gave the code a proper look and looks good. There's one UX thing:
The ellipse growing on hover is super nice, however by changing the ellipse size from an odd number to even (from 5 to 6 when not hovering), the ellipse will not appear centered to the cursor crosshair any more, not to a single pixel in the canvas.
Also, when moving a point, it is great that the GCP stays highlighted on the other canvas, however the 12 pixel marker on the active canvas kind of "obstructs" the view, while the previous 5 pixel marker was barely visible behind the mouse crosshair (a single pixel was visible in each quadrant).

I'd suggest to keep the 5 pixel ellipse when non hovering, use an odd number (13?) when hovering, and disable the hovering mode when a point has been picked, either in the current canvas or both canvases if that complicates things.

[nirvn]

@uclaros , we can disabled the hovered state when moving, that'll give us as much of a view of the canvas as possible.

With regards to the change from 5 to 6 point width, wasn't the previous size creating an issue whereas the marker wasn't aligned to the X,Y map canvas coordinate itself? When calling QgsGCPCanvasItem::paint, the QPointF( 0,0 ) there represents the center of the coordination location on the map canvas, by having a 5 x 5 ellipse marker drawn at QPointF( -2, -2 ), we end up with a marker with a misalignment of QPointF( 0.5, 0.5 ) against the coordinate it is representing:

So I guess then question becomes, is it better to be aligned to the crosshair cursor (I assume the shape varies across OSes, etc.) and be slightly misaligned with the map canvas itself, or vice versa. I thought map canvas accuracy was better, hence the change.