use crystal ball function for Z peak fit

[masonproffitt]

A relatavistic Breit-Wigner distribution would be the right function if the width of the peak was dominated by the actual width of the Z resonance, but at least in this case it's dominated by the detector dimuon invariant mass resolution.