Reference Issues/PRs

Fixes #21945.

What does this implement/fix? Explain your changes.

This PR adds a guide to the online documentation on how to craft a minimal reproducer oriented to bug report in scikit-learn. In particular it:

• Explains by giving short code examples on how to generate small synthetic datasets with the numpy random module, pandas, sklearn.dataset.make_classification, make_regression, make_blobs, or any of the scikit-learn loading dataset utilities;
• Explains step by step how to simplify the code and hyperparameters to find a minimal model that can reproduce the non-expected behavior;
• Explains that the effort made to craft a minimal reproducible code example will help remove any ambiguity in the English written description in the bug report and allow maintainers or other contributors to find a fix for the problem efficiently;
• Explains the markdown formatting for triple backticks with python or python-traceback qualifiers;
• Links this document to the contributing page;
• Adds the document to the index.

Any other comments?

This is still WIP, so any feedback on how to improve this document is always welcomed.