Multiple doc improvements

[ablaom]

Replaces #316. Closes #314.

This PR is chiefly about making doc generation more robust across all the examples. There are only small content tweaks.

This PR:

• Replaces all RNGs in examples with a manually seeded StableRNG to enforce reproducibility across Julia versions. (Some instability around RNG's caused the docs to crash after Julia 1.12 release).
• Dump RDatasets as a doc dependency, by using MLJ's built-in iris dataset. (Current version of RDatasets blocks using latest versions of MLJ)
• Remove the Project.toml/Manifest.toml files in each example folder, as these are not actually being used to generate the docs. Rather the /docs environment is used. Each example folder has README.md which now points users to that /docs environment.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.85%. Comparing base (108508e) to head (afb983e).
⚠️ Report is 18 commits behind head on dev.

Additional details and impacted files

@@            Coverage Diff             @@
##              dev     #318      +/-   ##
==========================================
+ Coverage   92.67%   94.85%   +2.17%     
==========================================
  Files          14       15       +1     
  Lines         519      525       +6     
==========================================
+ Hits          481      498      +17     
+ Misses         38       27      -11     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[EssamWisam]

Looks great to me. I remember experiencing confusion with the project.tomls. Thank you so much for this, I took a look and all looks good to me.

[EssamWisam]

Is removing this intended?

[ablaom]

Yes, in two cases I have intentionally removed the pre-executed versions of the notebooks. Literate.jl successfully creates executables for the others but fails on these (and this was true in the past for me). Previously, I would just execute by loading into jupyter but, as is typical in my experience of using Jupyter notebooks with Julia, there's some voodoo magic I have to figure out to get the latest julia kernel to load properly and I decided to give up this time around, because I don't like to maintain fragile workflows like this.

[EssamWisam]

Here also we have the unexecuted notebook only. I don't mind not having the other one though in both instances.

[ablaom]

Thank you @EssamWisam for the review!