Add support for random walk year effects with covariates. There are a few specific cases where covariates or the random year effect are not estimable. Examples are:
Import S3 methods from rstantools instead of rstanarm (#5)
Adjust calculation of year index values to better allow for missing time slices