docs: Restructure, add tour, add sidebar links to package docs

[aaron-hopkinson]

• Restructure the overview/package descriptions sections including:
  • Having a single description of Anemoi on the homepage (+ the package diagram)
  • Package descriptions moved to a separate page
  • Package dependencies moved to package description page rather than the homepage
  • Moved install instructions to their own page
• Added "A tour through Anemoi" page, linking to detailed docs following "user journey" discussions
• Misc small tweaks:
  • Add links to external packages (PyTorch, earthkit etc)
  • Small wording tweaks to improve clarity
  • Added alt-text to figures for accessibility
  • Links to other anemoi package docs on sidebar to improve navigability

---

📚 Documentation preview 📚: https://anemoi--26.org.readthedocs.build/en/26/

[VeraChristina]

This is great! Clear entry points to different parts of the ecosystems, and it generally looks much cleaner.

Two comments (that are probably out of scope for this PR):

• the figure on the landing page is difficult to read, and slightly biggerer figure/higher res would help a lot
• the "anemoi training getting started" link points to the user journey in the training docs, which is not really a quick start entry point. (I don't think there is a better point to link to currently, hence out of scope)

[aaron-hopkinson]

Thanks @VeraChristina

• Re: image on landing page – I just increased the size of the image. Hopefully that helps.
• Re: training quickstart – it looks like the Docathon training PR has been merged (docs(training): Docathon anemoi-core#201), so unless there's something else in flight, I suspect it's probably fine to merge this and then pick that up with the training docathon group. What do you think?

Re: images – The source currently has inlined raw HTML which I think would be better replaced with an rst figure directive, especially since the HTML currently has a <center> tag (which was deprecated back in HTML4, and is obsolete in HTML5, afaik). However, the embedded HTML does seem to use both object and img elements, which I think might be intended to support fallback from SVG to PNG versions of the images, which wouldn't be supported by the (arguably cleaner) rst version. However, I think all modern browsers support SVG, so we should probably just use this version. (Plus I don't know how well the fallback works – and using object over img seems less semantically clear, but I'm not a web developer...). Maybe one for a follow up PR, what do you think?

[VeraChristina]

Thanks a lot!
Re: image size -- yes, it's much more readable
Re: training quickstart -- I agree, let's pick this up separately with the training group
Re: raw HTML vs rst figure directive -- same here, let's pick this up separately