It boils down to: how much spacecraft resource is required by the attitude control method you propose using? And sometimes the mission's pointing requirements play a significant role.
Thrusters use propellant. Reaction wheels (and momentum wheels) use electric power. Spin stabilization uses neither, as long as you don't need to repoint the spacecraft — but any scientific spacecraft is going to need repointing, it's just a question of how often.
With reaction wheels, the amount of power used depends on the spacecraft's mass (actually, its inertial moment) and how quickly you need to turn. If you really need it, you can charge batteries with the electric power source (like an RTG) when the spacecraft's power demand is less than the source's power production capacity, and supplement that source with battery power when there's a high-power-drain period.
Still, though, reaction wheel turn rates are typically very slow due to the electric power requirements. If you need to turn really quickly, thrusters do better. But thrusters use up propellant. If you do a lot of turns, you can burn up a lot of propellant, and that adds more mass to the spacecraft at launch — it has to carry more propellant.
The decision about which approach to use usually involves engineers looking at the anticipated mission profile, notably the reorientation profile (how often? how far? how fast?) to see if one or the other would involve less resource use (such as mass) and they also weigh such considerations as how the different techniques mesh with mission requirements. For instance, if the science instruments require very accurate and exquisitely stable pointing, even if a thruster-based system might wind up being less massive than a reaction-wheel-based system, the reaction wheels are more accurate and stable and so you might choose that approach.
Spin-stabilized spacecraft typically have quite a bit of total angular momentum, so the small angular momenta of reaction wheels could precess such spacecraft very, very slowly. Thrusters generally are a better fit.
In Cassini's case, there were two considerations that drove them to reaction wheels: 1) they were doing a lot of turning through large angles, so using reaction wheels would save a lot of propellant mass; and 2) the exquisite resolution of Cassini's narrow angle camera (NAC) required exquisite pointing stability, better implemented with reaction wheels than with a separate set of small attitude control thrusters. By "separate set" I mean that Cassini had a thruster system for coarse attitude control and reaction wheel desaturation, but the minimum impulse you could deliver from them (from the shortest possible burn duration, called the "minimum impulse bit") wouldn't keep the pointing as stable as desired, so to do it with thrusters would require a separate, smaller set of thrusters with a much smaller minimum impulse bit to provide sufficiently accurate pointing.