feat: adding the hamming function - #3135

Description

Following the addition of hanning, this PR implements the Hamming window function (mlx.core.hamming). This further improves parity with NumPy's signal processing tools (numpy.hamming).

Implementation Details

• C++: Implemented hamming in mlx/ops.cpp.
  • Optimization: The angular factor 2π / (M-1) is pre-calculated as a scalar on the CPU before graph construction. This reduces the number of operations in the compute graph compared to a naive implementation.
  • Formula: Implements the standard Hamming window equation: $w(n) = 0.54 - 0.46 \cos\left(\frac{2\pi n}{M-1}\right)$.
  • Edge Cases: Handles M < 1 (empty) and M = 1 (returns [1.0]) explicitly to avoid division by zero.
• Python Bindings: Exposed via nanobind with explicit type signature (nb::sig) and full documentation in python/src/ops.cpp.

Test Plan

Verified locally against NumPy for numerical accuracy.

Verification script:

import mlx.core as mx
import numpy as np

M = 50
mx_hamming = mx.hamming(M)
np_hamming = np.hamming(M)

# Check for numerical closeness (float32 precision)
assert np.allclose(mx_hamming, np_hamming, atol=1e-6), "Mismatch with NumPy implementation"
print("✅ Verification passed against numpy.hamming")

Unit Tests:

Added test_hamming in python/tests/test_ops.py covering standard cases, M=1, and M=0.

Checklist

Put an x in the boxes that apply.

• I have read the CONTRIBUTING document
• I have run pre-commit run --all-files to format my code / installed pre-commit prior to committing changes
• I have added tests that prove my fix is effective or that my feature works
• I have updated the necessary documentation (if needed)