Remove Julia from conda dependency environment (but keep tests for esmvaltool install Julia) and add docs how to install from source

[valeriupredoi]

Description

Synopsis

• Julia from conda-forge is extremely old and unmaintained, needs removal (and gubbins needed to make it work) - this is absolutely necessary to do so since our env is getting impossible to solve
• Users can install their own Julia from source (instructions provided) if they with to run any Julia recipe (good luck with that!)

More info

Julia conda package is stuck at 1.10.4 which is an almost 9-month old version; they updated the base code to 1.10.8 https://github.com/JuliaLang/julia which includes a fix that permits us to free curl<8.10; freeing curl is becoming very important since that pin makes the env solving harder and harder, and holds back Python 3.13 too (not the only dep holding that though #3805 but an important one nonetheless); Julia conda feedstock appears to be rather abandoned https://github.com/conda-forge/julia-feedstock so am not sure how feasible and long term maintained it'll be to keep using the conda package; @bouweandela suggests we install Julia the "classic" way https://julialang.org/downloads/ and this exactly what we do here. A decision to remove Julia from our conda env has been taken at the ESMValTool workshop on May 15, 2025

Python 3.13

Is viable, and the only restrictions are the psy- packages:

-  psy-maps >=1.5.0
-  psy-reg >=1.5.0
-  psy-simple >=1.5.0

we will probably proceed to retire @schlunma 's recipe that uses those, then remove those packages from our env as well.

Julia recipes

They have been retired by @bettina-gier in #4045

• Closes Unpin curl #3758
• Closes Unpin (and remove) curl #3807
• Closes Bringing our environment to modern standards (also with removal of NCL) #3812
• Closes Remove ncl and its dependencies from main dependencies and add documentation how to install it #3853
• Closes Skip the Julia test if Julia isn't installed #3599
• Closes Julia fails when using a singularity container #3410
• Documentation added and created

---

Before you get started

• ☝ Create an issue to discuss what you are going to do

Checklist

It is the responsibility of the author to make sure the pull request is ready to review. The icons indicate whether the item will be subject to the 🛠 Technical or 🧪 Scientific review.

• 🛠 This pull request has a descriptive title
• 🛠 Tests run successfully
• 🛠 Any changed dependencies have been added or removed correctly
• 🛠 All checks below this pull request were successful

[valeriupredoi]

@bouweandela this is a viable way of installing and having Julia in our env that avoids installing (the very old) version on conda-forge; this way we can remove both curl and its pin and julia from our conda environment; note, however, that this introduces the issue we saw in the Python 3.13 PR whereby the environment becomes completely insolvable (because of: removing the pin on curl that holds the environment together, albeit with old dependencies, and because of NCL - if NCL is removed then the environment here solves very fast). I propose I implement a similar method to have NCL removed from our environment, and only installed in the CI tests, to reproduce what a user would do, should they need to run NCL diagnostics. Honestly, there's no other way around it. What do you think?

[schlunma]

Looks good to me, just some minor comments.

Cheers!

[valeriupredoi]

many thanks @schlunma - @bouweandela would you have 5 min have a look at this please? 🍺

[valeriupredoi]

this is pretty momentous actually - @bouweandela is sending regards to Julia, and OK-ed the merge offline, many thanks @schlunma for the review, bye, Julia 😁