Jupyter: Conda and Mamba

[ax3l]

Add hints on how to install complex software environments with mamba.
This overcomes some common conda issues.

[ax3l]

@proutrc @kincl @secondspass ready for review :)

[secondspass]

@ax3l can you provide some context on what problems conda has in adding new libraries? It's a little unspecific and not clear what potential problems mamba solves.

[ax3l]

@secondspass thanks for the feedback!

Yes, a main challenge is often that the conda default channels ship relatively old software, which are already present in the base environment. If people execute conda install ... the conda solver itself takes not only very long but also often finds undesired solutions, such as not picking up the latest release of requested software.

This is especially common when installing software from -c conda-forge on top of a non-empty, base environment with a couple of existing packages. mamba is a simple drop-in replacement for those tasks and is fully compatible with conda.

Some more info on the why's of mamba are in those blog posts:

• https://wolfv.medium.com/making-conda-fast-again-4da4debfb3b7
• https://wolfv.medium.com/mamba-development-news-29e32aaa8d6c

I can add more of those details in the docs, just did not know how detailed I should get. What do you think?

[secondspass]

That makes sense, can you write something like "since the conda dependency solver can sometimes be slow when adding packages and sometimes not pick up the latest release of a package, you can also use mamba..."? Just basic detail like that should be enough.