Install kernel dependencies with the -U flag

[DonJayamanne]

When install ipykernel, etc we should update the depednencies.
We have had plenty of cases where users had it installed, but versions were incorrect.
They had to uninstall & re-install.

For instance I had an issue with ipykernel in a conda environment.
Had an old version of ipython, hence when installing ipykernel, an old compatible version got installed. However that version of ipython is too old, similarly its possible the versio of ipykernel is also old.

The solution was conda install ipykernel --update-deps --force-reinstall
We might want to do the same for pip install ipykernel -U --force-reinstall

How to test/validate

• Install latest dev version of Python extension with the changes (here's the PR in Python ext Support for Jupyter to pass additional install args vscode-python#16133)
• Use conda & non conda
• Get notebooks working
• Uninstall pyzmq (a dependency),
• Now ipykernel is installed, but it doesn't work, hence re-installing ipykernel will not work
• Try now, and it should force re-install

[DonJayamanne]

We might also want to consider makign it possible to pass in the argument --force-reinstall.
We found at least in 2 occasions that this works.
hence we might have scenarios where we have -U as well as --force-reinstall.

TLDR - ensure we can pass more than just the one argument.

[DonJayamanne]

This works, see here #5689 (user tried this approach to get unblocked recently)

[DonJayamanne]

We should probably use --update-deps --force-reinstall only when first install of ipykernel fails.
Probably worth discussing with team.

[DonJayamanne]

Discussed solution:

• if ipykernel is available (pip list), then pass flags to Python to force update & force re-install

[joyceerhl]

I do get a prompt for conda and non conda environments, but for conda the install command fails. Rerunning the same command that the extension generates in the terminal doesn't work. I had to rerun the bare install command from Anaconda Prompt. Not sure if this is a VS Code integrated terminal issue.

Conda:

Conda install errors out:

Non-conda (pip):

[DavidKutu]

validated, in both the jupyter and python extensions