A Linear Theory of Multi-Winner Voting