Extract shared segment_wav into tts_dataset_utils.py - #16209

What does this PR do ?

• Extracted segment_wav into nemo/collections/tts/parts/utils/tts_dataset_utils.py.

• Added docstring to explain details of how segment_wav splits a waveform into fixed-length, zero-padded segments using a sliding window.

Why?

Inside scripts/ssl_tts, both make_supdata.py and ssl_tts_vc.py duplicate the same segment_wav function for splitting a waveform into PyTorch Tensor segments.

By moving this segment_wav into tts_dataset_utils.py, now these two files can import their shared segment_wav without risks of drifting from each other.

Tests

segment_wav now lives in tts_dataset_utils.py, so tests/.../test_tts_dataset_utils.py::TestTTSDatasetUtils gains two new test functions:

• test_segment_wav_pads_short_input — for shorter input that just needs padding.
• test_segment_wav_sliding_window — for longer input that triggers sliding window.

Before your PR is "Ready for review"

Pre checks:

• Make sure you read and followed Contributor guidelines.
• Did you write any new necessary tests?
• Did you add or update any necessary documentation?
• Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
• Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

• Refactor

Who can review?

@blisc