Leveraging the Quasi-Random Exposure of Swedish Men to Army Peers to Identify the Causal Effect of Marriage on Labour Market Outcomes