All the rocker r-base/r-apt based images already have the R recommended
packages pre-installed, see e.g. [0].

By installing the recommended packages also on the r-ver based images the
compatibility with packrat is improved, when it is used in a mixed mode
on top of different distribution with recommended packages pre-installed
and r-ver based images. Packrat does not include recommended packages in
its package library packrat.lock, if those are in the system library,
see also the pakrat docu page 5 in [1]. This leads to unresolved
dependencies and would require manually adding those also to the packrat
library.

The R installer already includes a mechanism to also install the
recommended packages, which is just activated by this change. The R
recommended packages are also versioned by the R version, see [2]-[6].

[0] r-recommended: https://github.com/rocker-org/rocker/blob/master/r-base/Dockerfile#L50
[1] https://cran.r-project.org/web/packages/packrat/packrat.pdf
[2] https://cran.r-project.org/src/contrib/${R-Version}/Recommended/
[3] https://cran.r-project.org/src/contrib/3.5.1/Recommended/
[4] R_PKGS_RECOMMENDED: https://github.com/wch/r-source/blob/trunk/share/make/vars.mk
[5] https://github.com/wch/r-source/blob/trunk/src/library/Recommended/Makefile.in
[6] https://github.com/wch/r-source/blob/trunk/tools/rsync-recommended

users can install libtcl8.5 & libtk8.5 downstream if packages need these (or the dev versions thereof)

Appears that R reads only Rprofile.site but not an Rprofile file in
R_HOME/etc.  However, R installation creates a Renviron in R_HOME/etc,
not an Renviron.site.  It appears also that RStudio copies these into
/etc/R, and renames both with the .site suffix.  If they do not exist,
or /etc/R does not exist, it does not end up reading Renviron at all.
Pretty confusing to me...

makes more sense to fix the build date when the R version is also set.  Leads to removal of images older than R-3.1.1,
since MRAN does not exist during 3.1.0 (nor did docopt). This also makes sure littler and the install2.r script are
available at all levels.