That's a good question! For performance road tires, we can make some inferences from Aerocoach's article here. For background, Continental put siping on the sides of their GP 4000 tire for purposes other than aerodynamics; I forget exactly why but aero was not at all on their minds. It was then discovered that the GP 4000 tested as surprisingly aerodynamic. The GP 5000s inherited the sipes. The GP TT is completely smooth.
That is, they tested (figures in weighted average drag at 45 km/h) that the GP TT is 1.7W slower than the GP 5000 in terms of aerodynamic drag. That difference is more than offset by the GP TT's lower rolling resistance, but it is not exactly an everyday tire. From that, I'd infer that for performance road tires, the difference in aerodynamic drag from the worst to the best probably isn't more than 5W. As I understand it, the sipes generate just enough turbulence that the airflow around the front of the wheel reattaches more easily to the back of the rim.
Now, the question is about tread, especially the knobs on MTB and gravel tires. As the knobs hit the wind, they do generate more turbulence than the GP 5000's tread does. This makes reattachment harder. This also interacts with the rim shape, by the way. If we are discussing MTBs where the rims aren't at all aerodynamically shaped (yet?), then perhaps the tread doesn't matter as much. The air isn't going to really reattach anyway, but there are many other more important areas to optimize.
Gravel is an intermediate case. Flo Cycling made a pair of 50+mm deep gravel wheels, and they discussed their aerodynamic considerations here. Gravel is a hard case because the amount and shape of tread can vary a lot - some tires are very knobby, some are just barely knobby, and there are courses where you'd run big slicks. Hunt is another company that has attacked the question as well, and they frequently publish white papers. I haven't searched their site, but I expect interested parties might find something.
