Both of them have been used successfully by many teams, and you can't go wrong using either for helping determine your robot's heading.
Honestly it is really hard to judge a gyro by its spec sheet - you always end up with apples vs. oranges (ex. the ADXRS has *excellent* tilt and vibration rejection specs, whereas the MPU-9250 can use its linear accelerometers for tilt rejection), and various manufacturers may be more or less conservative about the relationship between their specs and their actual median product performance.
When it comes to pure yaw performance, I'm a bit surprised that nobody has done a benchmark yet
(Put both gyros on the same robot, calibrate right before starting, drive around for 2 minutes, return and manually align the robot to exactly the same angle...repeat for N trials). I'd be shocked if there was a significant difference, to be honest.
Of course, there's no doubt that NAVX wins hands down in pitch, roll, linear acceleration, and direction of gravity measurement though.