Add DBSCAN clustering algorithm in machine_learning/ - #14851

Thank you for the review @zain-cs! You're right the already_in_another check was redundant after the noise reassignment. I've reordered the logic so the guard runs before the assignment, making the intent clearer. Please take another look when you get a chance!

@priya-sundaram-dev, please review vs.

• Implementation of Density-based spatial clustering of applications with noise (DBSCAN) ML Algorithm #11671

Reviewed both DBSCAN PRs side by side (#14851 vs #11671). My recommendation is to merge #14851 (this one) and retire #11671. Reasoning:

Correctness / algorithmic completeness

• This PR returns the standard DBSCAN contract: a per-point list of integer cluster labels with noise = -1, formed by expanding density-reachable neighbours into connected clusters. That's what a caller of dbscan(...) expects.
• Implementation of Density-based spatial clustering of applications with noise (DBSCAN) ML Algorithm #11671's perform_dbscan() only computes each point's ε-neighbourhood and then labels points core/border/noise for printing. It never assigns cluster IDs / joins density-connected components, so it doesn't actually produce clusters — it's closer to a neighbourhood report than a clustering.

Dependencies

• This PR is pure standard library. Good fit for the repo's "readable reference implementation" goal.
• Implementation of Density-based spatial clustering of applications with noise (DBSCAN) ML Algorithm #11671 pulls in pandas and matplotlib for what is a core algorithm, and its plot_dbscan() calls plt.show() inside a doctest.

Tests / style

• I ran this file's doctests locally — all pass (including the ValueError edge cases and the noise/single-cluster examples). Type hints and docstrings are clean.
• Implementation of Density-based spatial clustering of applications with noise (DBSCAN) ML Algorithm #11671 also carries an Author / LinkedIn block in the module docstring and uses type(...) is str rather than isinstance.

Two small optional nits on this PR (non-blocking):

1. The docstring advertises O(n²) — worth a one-line note that neighbours are recomputed for core points, so it's fine for a reference impl but not for large n.
2. get_neighbors is called again for every seed; caching would avoid recomputation, but I'd keep it simple for readability here.

Net: #14851 is the more complete and dependency-light implementation and is ready to merge; I'd close #11671 in its favour (with thanks to @tkgowtham for the earlier effort).