Hybrid Quantum-Classical Scheduling for Accelerating Neural Network Training with Newton's Gradient Descent