Benchmarking foundation models as feature extractors for weakly supervised computational pathology. — SciRadar