Add t-SNE (t-Distributed Stochastic Neighbor Embedding) algorithm #13432

Feature description

Enhancement: Add t-SNE implementation

Description:
t-SNE is a popular dimensionality reduction technique for visualizing high-dimensional datasets.
Currently, there is no implementation of t-SNE in the machine_learning directory.

Proposed Implementation:

• Implement t-SNE from scratch for educational purposes (no sklearn wrappers).
• Include clear docstrings, type hints, and doctests.
• Return the transformed dataset instead of plotting directly.
• Reference: van der Maaten & Hinton, 2008

Benefits:

• Allows users to learn how t-SNE works under the hood.
• Useful for educational projects and visualizations.