The Complexity of Learning Approval-Based Multiwinner Voting Rules