Prune dependency tree

[crsh]

Currently, papaja has some pretty heavy (indirect) dependecies, which we could strive to remove. Here are some ideas:

• gsl keeps giving users and our CI a hard time during installation. We rely on it indirectly through MBESS as a suggest. This should be solved shortyl, as MBESS has moved gsl to suggests.
• stringr imports stringi, which takes a long time to install. The following packages we import use stringr:
  • knitr: The package uses stringr quite a bit, but there is an open PR to remove stringr over the medium term.
  • rmarkdown: There is only one call to stringr, which can probably be easily removed.
  • broom: There are several uses of stringr, but they should, in principle, be easy to remove.
• broom: Has a heavy tidyverse dependency tree. We use it in some apa_print()-methods, but there we could switch to our alternative internal workflow (i.e. canonize() etc.) or use/support the output of a different package, such as parameters. Arguments for broom are the large number of supported models and it's widespread use, which makes it penatrable for independent contributors.

[mariusbarth]

With the newest version of arm (1.12-2), we do not depend on Hmisc and its imports, anymore 🎉