Multiply robust estimation of the causal effect of treatment in data subject to irregular observation times and confounding