Rethinking Tabular Data Understanding with Large Language Models