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

* Sort libraries, options and R packages alphabetically

As suggested by https://docs.docker.com/engine/userguide/eng-image/dockerfile_best-practices/#sort-multi-line-arguments

* Add packages and library as suggested by mdsumner

see https://github.com/rocker-org/rocker/issues/213#issuecomment-276968279

* Various changes in gdal compilation and structure

- Add dependencies to compile as homebrew gdal2 does
  https://github.com/OSGeo/homebrew-osgeo4mac/blob/master/Formula/gdal2.rb
- Add configure options to gdal to compile as homebrew gdal2 does
- change structure slightly so that debugging is easier
- add missing definition for $MRAN

* Add odbc and curl to gdal

* Cleanup to remove external libraries

* Dependencies changed

* Move to gdal 2.1.3 and add KML support

- add some structure to dependencies

* Add LD_LIBRARY_PATH

* Fix HDF5 support

Error in detection HDF5 libraries in 2.1 branch - see
http://permalink.gmane.org/gmane.comp.gis.gdal.devel/44474 for info.

* Fixed Typo for HDF5 library location

Thanks Edzer

* use debian build-dep to get dependencies
* include library path

Add GROUPID and UMASK options to RStudio image

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...

provides better relative timing to build each image, also provides more granular error checking without inspecting full log

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.