Supporting JupyterHub singleuser 3.0.0

[manics]

I had a look at what's required for this to bump jupyterhub-singleuser to 3.0.0 (jupyterhub/binderhub#1544 (comment)).

JupyterHub 3.0.0 requires Python 3.7+

Python 3.7, 3.8, 3.9, 3.10 environments

This is easy, just update

repo2docker/repo2docker/buildpacks/conda/environment.yml

Line 11 in d7be04e

- jupyterhub-singleuser==1.5.0

and run freeze.py to update Python 3.7, 3.8, 3.9 and 3.10 environments.

Python 2.7 environment

Also easy, since the kernel environment is decoupled from the notebook/JupyterLab environment.

Python 3.5 environment

Not automatically updated:

repo2docker/repo2docker/buildpacks/conda/environment.py-3.5.yml

Lines 1 to 2 in d7be04e

	# py3.5 is not being auto-generated from this file anymore
	# update environment.py-3.5.frozen.yml directly

Stuck on a very old version, too old for 3.0.0:

repo2docker/repo2docker/buildpacks/conda/requirements.py-3.5.pip

Line 8 in d7be04e

jupyterhub==0.9.4

Python 3.6 environment

Not automatically updated:

repo2docker/repo2docker/buildpacks/conda/environment.py-3.6.yml

Lines 1 to 2 in d7be04e

	# py3.6 is not being auto-generated from this file anymore
	# update environment.py-3.6.frozen.yml directly

Stuck on an old version, too old for 3.0.0:

repo2docker/repo2docker/buildpacks/conda/environment.py-3.6.lock

Line 131 in d7be04e

https://conda.anaconda.org/conda-forge/linux-64/jupyterhub-base-1.4.2-py36h5fab9bb_0.tar.bz2#01e91ea234e632cf4ee3284a338a679d

Options

• Separate Python 3.5 and 3.6 environments from the notebook/JupyterLab environment, similar to Python 2.7
• Separate all Python environments from the notebook/JupyterLab environment. If we're going to migrate 3.5 and 3.6 environments to the decoupled setup, it probably makes sense to do it for all Python versions to make things easier in the future, and also solves some potential package conflicts: Fresh Python kernels for reliable builds. #741
• Separate all environments for all languages from the notebook/JupyterLab environment. Architecturally I think this makes sense, but is more work: Treat jupyter as an editor (decouple jupyter and user environments) #868