4- and 2- inch PM dataset and one 20_6 million LUT

The 4-inch and 2-inch dataset are generated using random mover poses with single square coils. The sample sizes are 8 million and 9.6 million for the 4-inch and 2-inch datasets, respectively. The input and label data are min-max normalized to range between zero and one. The 20.6 million dataset file is created using the same wrench model (GT model), which has infeasible mover poses. The infeasible mover poses are scruntinized for the 4-inch and 2-inch dataset to make sure the PM is above the top surface of the coil. The wrench results for the infeasible locations produced large results due to zero-divide-zero operation. The issue are fixed using another numerical method which is slower than the GT used in the study.