Describir: Stacking Ensemble Learning with Geomagnetic and Meteorological Features for Atmospheric Electric Field Forecasting