The short answer to your first question is "the power savings from using ceramic bearings compared to good steel bearings is almost zero." The short answer to your second question is "yes, it is possible to measure the difference but it's not easy." The longer answer, and the support for the shorter answers is below.
First, however, it depends a little on whether you're talking wheel bearings, or bottom bracket bearings, or both. If you're interested in measuring the difference, the way you measure differs for the two. BB bearing drag would show up in differences in drivetrain efficiency, while hub bearings would show up as a component of rolling resistance. The difference, of course, is that drivetrain losses only occur when you're pedaling, not while you're coasting, while rolling losses occur as long as you're rolling, whether pedaling or not.
You can measure differences in no-load hub bearing drag by holding the wheel up in the air with a fork, using a home-made speed sensor jury-rigged with a reed switch connected to a digital voice recorder. You spin the wheel up, then record and calculate the difference in timing "clicks" as the wheel decelerates. Swap bearings, lather, rinse, repeat.
However, this would be in a no-load setting and you might think differences in bearing drag might be amplified in load-bearing situations. If so, you'd have to test that way. You could do it in the lab but it's also possible to do yourself, in the field, with the jury-rigged speed sensor.
First some background. The rolling component of drag is typically measured by the coefficient of rolling resistance, Crr (the coefficient of aero drag is Cd, which is usually multiplied by the front surface area, A, and described by the combined term CdA). Crr can have a speed-dependent component but at cycling speeds this is negligible so we can assume Crr is constant. If so, the power demanded for moving a bike and rider at a given speed is pretty well known and understood: rather than go through it all here, I'll point to a (um, brilliant) method that allows you to estimate CdA and Crr which can be found here.
Now, if you do all of this, you'll find that for a typical bike, the total overall Crr from all sources (tires, tubes, hub bearings, and typical "smooth" road surface) of around .005. The Crr for really good tires on that same surface might be .0045; for really lousy tires the Crr might be .006; and for a really rough surface the Crr might go as high as .01. And, some context if you look at the power equation linked to above, you see that Crr scales exactly like gradient in its effect on power so when the Crr is .01, the effect on power is exactly like climbing a 1% grade. Likewise, an increase in Crr of .001, due either to a different tire or to putative bearing differences, is just like climbing a grade 0.1% steeper.
Here's one more fact, also derivable from the link above: a rule of thumb is that at 25 mph on a flat hard surface, a cyclist in good aero position needs around 250 watts (an exceptional aero position is achievable with under 200 watts). At 25mph, a difference in Crr of about .0005 is roughly equivalent to a difference in power of 5 watts. That is, a difference between a good tire and a very good tire is worth about 5 watts at 25 mph.
If you have done any field-testing of bicycle drag you'll realize that 5 watts difference in rolling resistance is pretty noticeable. Riders report that the difference between using ceramic and steel hub bearings is not noticeable.