Add support for launching quantized weights on 310p NPU like Atlas 300i duo cards. There is none backends which supports quantization for 310p except mindIE(only INT8 with special format convertion). Quant support will make this card best value for local inference.