Add dataset processing script to align clean and noisy speech via STFT

[aryanKaga]

Summary

This pull request adds a script that processes clean and noisy .wav files in the MS-SNSD dataset, extracts their STFT magnitude spectrograms, aligns them by time, and stores them in .npz format for efficient machine learning training.

Features

• Matches noisy and clean speech files using clnsp ID extraction
• Applies convert_to_magnitude() from stft.py
• Crops clean/noisy STFT matrices to the shortest length

Let me know if you'd prefer a different format or integration method. Happy to update the structure

[aryanKaga]

@microsoft-github-policy-service agree