Great question! I doubt anyone has tried, but I found the question really interesting, so here's my long-form thinking and a little back of the envelope math on the question.
The sensor can stream depth frames at 30 frames/second for 640x480 and 60 frames/second at 320x240. Given the nature of the problem, I think you'd probably willingly sacrifice the resolution for a higher frame rate, so lets assume 60 frames/second.
According to a quick google search, the average exit velocity of a baseball in a high school game in roughly 70 mph. Based on personal experience, the ball goes less far when hit off a tee than when hit against a live pitch, but not dramatically so. So lets estimate that a high school player hitting off a tee can average 60mph in exit velocity off a tee. 60mph = 88 ft/second, or roughly 1.5 feet of expected ball flight per 60th of a second, aka 18" per depth frame.
To calculate exit speed and launch angle, you need at least one depth frame of the ball on the tee and at least one frame of the ball in mid-air (shortly after being hit), the more frames the better. Since the person is hitting off a tee, you can probably position the camera pretty close to the tee behind the batter, lets estimate 3 ft way from the original position of the ball to the back and to the side.
Eventually, the ball will get far enough away from the camera that there will be considerable error in its distance. We can track that error with this chart. So initially, at 3 ft ~1000mm = +/- 3mm error. Pretty damn good. We want to try to calculate the point at which that error is too large for our velocity calculation to be meaningfully precise. Lets assume calculation simplicity that the person pulls the ball at just the angle you are to the side, so the ball is moving directly away from the camera (could be at an upward trajectory if the sensor is below the original tee height).
At 7.5ft of ball flight you are 10.5 ft from the camera, corresponding to a depth error of roughly +/- 30 mm or 3cm. We would expect this to be the 5th depth frames of ball flight and to have happened in 5/60th of a second. That 3cm error would correspond to 3cm / (5/60 second) = 36 cm/sec which is a little under 1 mph (44.7 cm/s). Once you get more than say 1.5 mph you're starting to get worse than a radar gun, but we're also running into the outer edge of the chart here. So at 3500 mm, we expect 3.85 cm error, which is ~11.5ft, or 5.67/60 seconds, which comes out to 40.7 cm/s, still < 1mph error. I'd say that's a pretty solid exit velocity calculation.
Now lets briefly consider launch angle error. A baseball is ~7.5 cm in diameter. At 11.5 ft (350 cm) away from the camera,
the ball is going to be really tiny on a 320x240 resolution frame. A ball hit at, lets say a 30 degree angle would expect a height component of sin(30 deg) = .5* 350cm = 175cm and a ball hit at a 29 degree angle would expect a height component of sin(29 deg)*350 = 169.7cm. Throw in a couple cm in depth error and you could be looking at the entire 7.5cm diameter of the baseball corresponding to just a 1 degree error in launch angle. I.e. if you can pick up that ball at all in depth, it doesn't matter if its the middle of the ball or the very edge, you're going to get a pretty accurate launch angle.
So, while I was typing this, lelloii pointed out the sunlight issue, although that wouldn't be a problem in a dome or an indoor batting cage (although there this would only really work for balls hit up the middle so you can see them travel at least 5-10 ft). Also if you just want to do exit velocity without launch angle, a radar gun is probably just as good and way easier.
All this taken into consideration, I actually do think this could work, and I kind of want to try it sometime. You might be familiar with MLB's Statcast system, but it's basically trying to do the same thing with cameras that are much further away and are probably just really high frame rate / resolution rgb cameras. Knowing the exact size of the ball and it being round really does make it a very solvable problem with rgb camera as well.