std::vector< int8_t > pop_n_rows_from_merged_heaps_gpu(Data_Namespace::DataMgr *data_mgr, const int64_t *dev_heaps, const size_t heaps_size, const size_t n, const PodOrderEntry &oe, const GroupByBufferLayoutInfo &layout, const size_t group_key_bytes, const size_t thread_count, const int device_id)