consider reading .bar file for greater precision

Less relevant than it used to be (I think `.par` files are now written in greater precision than they used to be), but the full floating-point precision of estimated parameters is available in the `.bar` file: we should consider reading that if it's available (see `glmmADMB` package for an example)

No file or test is named in the issue. Start by locating the existing .par parameter-reading entry point, compare the available .bar precision with the glmmADMB example, and define completion as using .bar data when available without breaking the current path.

Written by the indexing model from the issue text.