could we optimize interactions by setting signals instead of re-writing the spec?

[jonathanzong]

currently, in the interactions code, when you change an Interaction definition it will re-render the whole visualization and write out different signals into the spec. this was easier to implement, but means that changing the interaction definition is somewhat costly, and there's often a couple seconds of delay afterward. i wonder if it would be more efficient to write out all the signal specs one might possibly need and then set signal values that switch interactions on/off using the pattern set up for mark encodings in signalActions and ctrl/signals.ts. ideally, this would mean changing the interaction definition would no longer require a re-render.