Machine learning enables prompt prediction of hydration kinetics of multicomponent cementitious systems