How do you conduct regression testing during release cycles to ensure that new code does not break existing functionality?