Into the LAIONs Den: Investigating Hate in Multimodal Datasets