Verify README's installation instructions

1. Don't suggest to set `JUPYTER_ENABLE_LAB`

The docker-stacks documentation states...

-e JUPYTER_ENABLE_LAB=yes - Instructs the startup script to run jupyter lab instead of the default jupyter notebook command. Useful in container orchestration environments where setting environment variables is easier than change command line parameters.

But if the startup script then starts jupyter using jupyter lab instead of jupyter labhub (or jupyter-labhub) we run into trouble, right?

NOTE: I'm already updating the zero-to-jupyterhub-k8s docs to not mention this environment variable.

2. Suggest either `jupyter labhub` or `jupyter-labhub`

Let us simplify and suggest using one or the other, assuming there is not difference. Speaking of which, is there a difference?

JupyterLab installs a script called jupyter-labhub and I think it does the same thing as running jupyter labhub. I wonder if there is a technical reason for writing ...

c.Spawner.cmd = ['jupyter-labhub']

instead of ... ?

c.Spawner.cmd = ['jupyter, labhub']

I figure there can be, related to having a single word as command and recall reading something about that.

For reference, here is an excerpt from jupyterlab/setup.py:

# Force entrypoints with setuptools (needed for Windows, unconditional
# because of wheels)
setup_args['entry_points'] = {
    'console_scripts': [
        'jupyter-lab = jupyterlab.labapp:main',
        'jupyter-labextension = jupyterlab.labextensions:main',
        'jupyter-labhub = jupyterlab.labhubapp:main',
        'jlpm = jupyterlab.jlpmapp:main',
    ]
}

https://github.com/jupyterhub/kubespawner/issues/227 - Danger of setting c.Spawner over c.KubeSpawner ?

@blink1073 clarified on Gitter that...

Technically when launching "Jupyter foo" you create one process that spawns "Jupyter-foo", so it is more efficient to run "Jupyter-foo" directly

Conclusion: We should suggest Spawners to be configured to use jupyter-labhub.

jupyterhub / jupyterlab-hub