Set PYTHON_JULIAPKG_PROJECT for Docker

Being in a Conda environment is effectively equivalent to setting:
   export PYTHON_JULIAPKG_PROJECT="$CONDA_PREFIX/julia_env"
The JuliaPy PythonJulia stack (and juliapkg in particular) tries to automatically detect whether it's running inside a Conda or Python virtual environment. 
But when building a Docker image we're not _exactly_ in a conda environment
we're kind of faking it, and so it doesn't detect it and put stuff
in the right place. So in the docker context we have to set it explicitly.
Let's make it explicit everywhere.

Author: rwest

Dockerfile

@@ -88,6 +88,9 @@ RUN make
 # setting this env variable fixes an issue with Julia precompilation on Windows
 ENV JULIA_CPU_TARGET="x86-64,haswell,skylake,broadwell,znver1,znver2,znver3,cascadelake,icelake-client,cooperlake,generic"
 ENV RMS_BRANCH=${RMS_Branch}
+# Usually this is set automatically, but we're not actually running
+# in an active conda environment when building the Docker so we need to set it manually
+ENV PYTHON_JULIAPKG_PROJECT="/miniconda/envs/rmg_env/julia_env"
 RUN source install_rms.sh
 
 # RMG-Py should now be installed and ready - trigger precompilation and test run

install_rms.sh

@@ -46,10 +46,12 @@ echo "Current conda environment: $current_env"
 conda env config vars set JULIA_CONDAPKG_BACKEND=Null
 conda env config vars set JULIA_PYTHONCALL_EXE=$CONDA_PREFIX/bin/python
 conda env config vars set PYTHON_JULIAPKG_EXE=$(which julia)
+conda env config vars set PYTHON_JULIAPKG_PROJECT=$CONDA_PREFIX/julia_env
 # Also export for current shell/session (needed for Docker/non-interactive use)
 export JULIA_CONDAPKG_BACKEND=Null
 export JULIA_PYTHONCALL_EXE="$CONDA_PREFIX/bin/python"
 export PYTHON_JULIAPKG_EXE="$(which julia)"
+export PYTHON_JULIAPKG_PROJECT="$CONDA_PREFIX/julia_env"
 
 conda install -y conda-forge::pyjuliacall