Bag of Tricks for Multimodal AutoML with Image, Text, and Tabular Data