Add pixi-kernel to base environment to bypass having to setup custom kernels of bespoke virtual environments

[matthewfeickert]

Previously, if one wanted to make an ipython kernel for a bespoke virtual environment that the coffea-casa JupyterLab had created in it one would have to do something like

# . /path/to/venv/bin/activate
$ python -m ipykernel install --user --name="<kernel-name>" --display-name="<kernel-display-name-from-launcher>"

and then after a few minutes the kernel would appear in the launcher. This however requires curation of all kernels for all environments you make (though this isn't a big deal in general as you're coming to coffea-casa to use the environments already created for you).

Using pixi-kernel one can do the same thing by just (on coffea-casa) installing it with

python -m pip install --upgrade --user pixi-kernel

as prescribed for situations where users don't control the JupyterHub/JupyterLab instance/environment.

To avoid this pattern (as --user installs should be avoided in general) would it be possible to normally install pixi-kernel (available from both PyPI and conda-forge) into the same base environment that contains JupyterLab? As seen in renan-r-santos/pixi-kernel#51 (comment), this allows for users to select the kernel they should use from the a dropdown but all of this only uses a single kernel square (Python (Pixi)) from the launcher.

Similar to Issue #437, if people agree on this, and I'm told what Dockerfiles are actually getting used, I could do this myself.