Efficiently Computing Cosine Similarity: A Performance-Critical Task Using Vectorized Computations with NumPy and SciPy