Into the LAION's Den: Investigating Hate in Multimodal Datasets