update.PKNCAresults errors on ordered-factor group columns and emits per-subject warnings for unchanged groups

[billdenney]

Found while documenting update() (which is a great feature — the book now covers it):

1. With datasets::Theoph (whose Subject is an ordered factor), update(o_nca, data = <modified conc>) fails in an internal dplyr join with an incompatible-ordered-factor-types error after changed-group filtering. It works when the subject column is character.
2. With automatic per-subject intervals, updating one subject's data emits a "Subject=N: No concentration data" warning for every unchanged subject (the dose side keeps all rows while the concentration side is filtered), which reads like something went wrong when nothing did.

The v01 vignette's update examples previously displayed nine of these warnings unexplained; the pending vignette PR adds an explanatory sentence, but quieting the false alarm at the source would be better.

[billdenney]

Make sure that ordered grouping factor columns retain their order and levels.