Welfare Measures When Agents Can Learn: A Unifying Theory