To have much effect on a regular plane the mass will either need to be quite a large portion of the model's weight or it will need to shift over good long distance within the model. You don't say what size of model you're considering but when you're talking about sail winches I'm thinking that it'll be up around a 40 to .60 size or electric model equivalent.

This suggests a 5 to 7 lb model weight. To shift the CG over about a 10 to 15% range of the chord to go from "mild to wild" you're looking at likely needing to shift a 7 or 8 oz mass over probably 5 or 6 inches of travel. Or a 4 to 5 oz lump over a longer distance such as around 10 to 12 inches.

I'm thinking that with something like this you might be better off with a weight on a guide of some sort and a closed loop "clothes line" setup with pulleys at each end and a geared motor turning a drum with a few wraps of a strong line around it. This can actually be made from a modified regular servo by removing the feedback pot and driver board and run it from a switch on a regular servo and limit switches on each end.

The jack screw idea is OK to but not the way you're considering it where you run the screw from a regular servo. You won't find any such thing as a 1/2 or 1 thread per inch "screw". The angles are such that it would take a massive torque to turn such a thing. Instead you're again looking at making up a custom "servo" or running a motor to a gear which is threaded internally to run on a normal length of small threaded rod such as 10-24. The power needed and the gearing down would come from the number of turns you need to travel the length of the rod with that fine a thread.

There's a few threads around that detail the amount of mass and distance needed to shift the CG of a model. Find one of those or look up the equations needed to calculate the mass torques and how it affects the weight of the model. That and running the model numbers through a CG calculator will give you a feel for how big a mass it needs to be and how far it has to move.

"Wild" actually implies being able to shift the CG to a slightly negative setting at a few % of the MAC behind the Neutral Point. The model will need to be constantly corrected but flying an "unstable" model up to a point is no worse then the tiny steering corrections we make while driving. It'll get worse as you approach a point where it's -3 to -5% for the Stability Margin but you won't know where the limit really is until you try it. There's a fair number of folks that fly with only 1 or 2% positive stability and don't find the models hard to control. And the flat foamy flippy models that do all the power hovering are commonly running negative stability margins of one amount or another since it makes the models hover in a more stable manner. Yet they don't have any problem flying around "normally" and being controllable.