When reordering incidence, also change the priors in estimate_advantage

[sangeetabhatia03]

Because of the reordering of incidence in estimate_advantage, users can get erroneous results if there is not sufficient data to inform the posterior distribution of epsilon. Say the prior has mean 5. and the true transmission advantage is 2. If we have enough data to modify the posterior, then we would estimate a disadvantage of 1/2 with reordering. As a final step, we fix everything and return 2 back. However, if there is no data to inform the prior, we don't return 1/5 but rather 5 (prior mean) and then as a final step, that is swapped to 1/5. The issue could be resolved perhaps by working with 1/prior mean when reordering.