Fix block_covariance remainder window handling

[rajanandsingh]

Summary

• Align block count with idx_range in block_covariance to avoid reshape errors when samples are not divisible by window.
• Add regression test covering the remainder case.

Repro

import numpy as np
from asrpy.asr_utils import block_covariance
data = np.random.randn(25, 26302)
block_covariance(data, window=100)

Tests
pytest