Don't use the main stator coils to sense RPM. Put a small neo on one of the rotors and use a coil to pick up the signal. Use a small fermite core in the coil to saturate it at a reasonable voltage level at all meaningful speeds. No harmonics and saturation will limit the voltage rise.
Alternatively use an optical pickup set up like an encoder. You don't need the precise angular position of the rotor so it can be sloppy.
Using the power signal as an input to control logic is a bad, bad idea. I'm assuming that you want RPM in digital for use in a load controller. The harmonics that are in the line already are nothing compared to the feedback from using the power waveform directly in your controller.
The inductance of the stator coils combined with any load change is going to cause voltage spikes that will mess with your logic and potentially fry it. If the controller to doing PWM dump control or charge management the load changes will feedback into the input and ring chaotically.