Another idea. Dead simple.
If I stir my cup of coffee the water keeps turning for some seconds after. In reverse, if I rotate my cup a bubble on my coffee sits still for the first few rotations.
Use can of juice. Drill a tiny hole in one end, and top it up with water, using a syringe. Block the hole with tape, or a small wad of the blue goo teachers use to put posters on wall.
Now you have something that is fairly heavy, -- lots of M * g. But the water inside won't spin with the can -- or not much.
You may be able to tweak this a bit with liquids other than water. You want to lower the viscosity of water so there is less drag. Your initial experiments with this could be done with jars with screw caps. Easier to change the liquid.
Methanol has about half the viscosity of water. There are materials that are lower -- benzene, acetone, chloroform, but they are not things that I'd want to slop around trying to get them into a can.
Soap solutions may be better.
You can get a quick test of viscosity by dropping something through a tall graduated cylinder of the liquid. The speed of drop depends both on the weight of the object, (Mg is hte downward force) the volume of the object (rho V is the buoyant force on the object, where rho is the liquid density) and the drag of the object moving through the liquid. Water vs water with soap won't have a measurable change in density. I vaguely recall that methanol has a density of around .95
This method will have slightly worse performance than the next one, but it's quick. Worst case, bring a can of apple juice.
Another idea.
My first idea was to put a rod in the middle. But if you do that, then the rod is going to rotate, and it will be fussy to get it exactly centered.
Instead run a thread through the middle. The thread is anchored at each end of the can. Run the thread through bunch of large (1") nuts, or fishing sinkers. The thread doesn't need to be tight, but it has to keep the weights off the ground. Fishing line may work better than thread.
The principle:
One of two things happen: The thread twists up as the can rolls, but the weights just hang.
Or the thread rotates, rotating the nuts at a slower speed proportional to the ratio of the nuts inner diameter and the thread.
This might be converted into a rubber band motor. The idea here is similar to the flywheel/gyroscope , but where the flywheel used the kinetic energy of the wheel slowing to speed up the can, the rubberband accelerates a rotating something to spin the can faster.
Mechanically it would work similarly: A rotor that is supported by the ends. The shaft is wrapped with a rubber band under tension.