Investigating Deep Learning-based Strategies to Cope with Label Noise in Satellite Data Using the Example of African Non-Forest Trees