feat: add outliers_nb_target to approximate_distribution() for target-based min_similarity [stacks on #2519]

[pidefrem]

What does this PR do?

feat: add outliers_nb_target to approximate_distribution() for target-based min_similarity

approximate_distribution() requires manually specifying min_similarity. Users must iterate to find a value that produces an acceptable number of documents with zero topic distribution.

Changes:

Add an outliers_nb_target parameter (int): auto-searches for the min_similarity value that achieves the target number of zero-distribution documents using binary search.

# Before: manual trial-and-error
distributions = topic_model.approximate_distribution(docs, min_similarity=0.1)

# After: specify the target directly
distributions = topic_model.approximate_distribution(docs, outliers_nb_target=50)

Performance optimization: Refactors the internals so the expensive similarity matrix is computed once and reused across all binary search iterations:

• compute_similarities_for_approximate_distribution() — compute once
• compute_topic_distributions_for_approximate_distribution() — apply threshold, iterate

Cannot set both min_similarity and outliers_nb_target. Backward compatible: defaults to None.

Fixes #2502

Before submitting

• This PR fixes a typo or improves the docs (if yes, ignore all other checks!).
• Did you read the contributor guideline?
• Was this discussed/approved via a Github issue? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes (if applicable)?
• Did you write any new necessary tests?

---

⚠️ Draft — stacks on #2519. This branch includes the reduce_outliers auto-threshold commit from #2519 (and its upstream prerequisites) as [prereq] commits, so the diff here shows that work too. Once #2519 merges, I'll rebase onto master and the prereq commits drop out, leaving only the outliers_nb_target change for approximate_distribution(). Marking ready for review after #2519 lands.