Avoid importing IPython if notebook cells do not contain magics

[neutrinoceros]

Description

follow up to #3748: completely avoid import overhead in jupyter_dependencies_are_installed. This function now takes <0.2ms regardless of the result, while the current implementation takes >200ms on first call when the result is True.

Checklist - did you ...

• Add an entry in CHANGES.md if necessary?
• Add / update tests if necessary?
• Add new / update outdated documentation?

[hauntsaninja]

Hmm, I thought about this, but when this returns true we'll end up importing these anyway right? In which case this introduces overhead

[hauntsaninja]

I guess we can save if there isn't a cell magic in the notebook

[neutrinoceros]

TBH I didn't look at how the function was integrated in the rest of the control flow. From an external point of view I thought it might be preferable to decouple introspection and actual imports, especially if the added overhead is 3 orders of magnitude bellow the import cost, but your call ! :)

[github-actions]

diff-shades reports zero changes comparing this PR (0afa13c) to main (0e26ada).%0A%0A---%0A%0AWhat is this? | Workflow run | diff-shades documentation

[ichard26]

The idea sounds reasonable and the changes look good. Please address failing lint CI though.

[hauntsaninja]

Yeah, sorry, lint is failing because of click stuff again that's unrelated to these changes, #3791 to fix

[ichard26]

Thanks @neutrinoceros and @hauntsaninja!