Installing PyPlot in a robust way

Installing PyPlot with Pkg.add("PyPlot") often fails. In this case it is recommended to install PyPlot in the following way:

1. Install a Python 3.x distribution that contains Matplotlib.
   Recommended: Anaconda distribution.
   Advantage: very robust;
   Disadvantage: > 3 GByte memory needed;
   ModiaMath is based on the Python 3.x version of Matplotlib where some keywords are different to the Python 2.x version.
   On Windows 10, either provide a julia_start.bat file with the following content (and add <path-to-julia-installation>\bin to the path environment variable):

   call <path-to-Anaconda3>\Anaconda3\Scripts\activate.bat
   julia
   

   and always start julia by calling julia_start,
   or (my preference), add additionally the following directories to the path environment variable:

   <path-to-Anaconda3>\Anaconda3
   <path-to-Anaconda3>\Anaconda3\Library\mingw-w64\bin
   <path-to-Anaconda3>\Anaconda3\Library\usr\bin
   <path-to-Anaconda3>\Anaconda3\Library\bin
   <path-to-Anaconda3>\Anaconda3\Scripts
   <path-to-Anaconda3>\Anaconda3\bin
   <path-to-Anaconda3>\Anaconda3\condabin
   

   and start julia by calling julia

2. Include the path to the Python executable in your startup file (Julia 1.0: <path-to-user>/.julia/config/startup.jl):
   ENV["PYTHON"] = joinpath("<path-above-Anaconda3>", "Anaconda3", "python.exe")

3. Start Julia, give the command ENV["PYTHON"] in the REPL, and check whether the path is correct (if you made a typo in the startup file, Julia might use another Python executable and PyPlot might crash Julia).

4. If you have used a different Python installation before, execute the command ]build PyCall (or using Pkg; Pkg.build("PyCall"), exit Julia and start Julia again.

5. Install PyPlot via ]add PyPlot (or using Pkg; Pkg.add("PyPlot"))