Improving Model Chain Approaches for Probabilistic Solar Energy Forecasting through Post-processing and Machine Learning