Analytical and Machine Learning-based approaches to estimate the steady-state temperature limit on the surface of Cu powder beds when heated by a concentrated laser energy source