Estimating the causal marginal effect of a continuous exposure on an ordinal outcome in the presence of confounding and covariate-driven monitoring times