Description

In spark 3.3， decimal * decimal * decimal , the precision will be increased at the same time, will use checkoverflow to decrease scale. Some times, decrease by int256, this is an inefficient operation.