Accelerating Binary and Mixed-Precision NNs Inference on STMicroelectronics Embedded NPU with Digital In-Memory-Computing