Readthedocs configuration/builds: revert to miniconda before miniforge is available

[valeriupredoi]

Description

Related to #3779 - while RTD implement and make miniforge3 available, we should use something else other than mambaforge to avoid the increasingly frequent "brownouts".

Docs with expected failure from the config changes https://readthedocs.org/projects/esmvaltool/builds/26029832/

---

Before you get started

• [x] ☝ 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.

• [x] 🛠 This pull request has a descriptive title
• [ ] 🛠 ~All checks below this pull request were successful~ docs fail with usual issue

[bouweandela]

The documentation build is failing. Is this ready for review?

[valeriupredoi]

nope, I need to reinstate that silly metadata package again - this all feels like going backwards, unless we have other options (that I can't think of)?

[bouweandela]

Solving the environment with this setup takes about 350 seconds (miniconda-latest), while with the current main branch (mambaforge-22.9) it takes about 600 seconds. How about deleting all the commented-out stuff, merging this, and forgetting about miniforge3? Or would you expect further improvements from that?

From this blog post, it looks like mamba is already the default solver for conda: https://www.anaconda.com/blog/a-faster-conda-for-a-growing-community

[valeriupredoi]

That'd what I thought too about mamba, but polling the RTD env for mamba returns executable not found - perhaps the mamba solver is used, but not the package with its executable? At any rate, I'd like to see what miniforge3 can do for us, that's why I'd leave all those comments in? Cheers buds 🍺

[bouweandela]

Would it work for you to move the commented out stuff to an issue instead? Leaving commented out stuff lying around really makes a mess and will only confuse future maintainers (probably ourselves..)

[bouweandela]

perhaps the mamba solver is used, but not the package with its executable?

That sounds plausible

[bouweandela]

This link seems to confirm that: https://github.com/conda-incubator/setup-miniconda?tab=readme-ov-file#example-6-mamba

[valeriupredoi]

@bouweandela I have just cleaned up the commented out bits in 2352ebca9c66b62898397985996bcb270704edec - a good call indeed to (temporarily, perhaps) remove all that guff :beer:

[valeriupredoi]

I've also updated the OS to ubuntu-lts-latest as per https://docs.readthedocs.io/en/stable/config-file/v2.html#requirements-file - who'd have known I can update an OS, somewhere :laughing: