[whisper] Preserve per-file output names in the CLI

[saichowdary007]

Summary

Fixes #1441.

When the mlx_whisper CLI receives multiple audio paths without
--output-name, derive the output basename independently for each input. The
previous loop mutated output_name after the first file, so every later
transcription overwrote the first output. Explicit --output-name behavior and
stdin handling are unchanged.

Testing

• PYTHONPATH=. .venv/bin/python -m unittest test_cli -v
• .venv/bin/pre-commit run --files whisper/mlx_whisper/cli.py whisper/test_cli.py

The regression test mocks transcription and verifies that two input paths are
written as first.txt and second.txt.

Implementation and test preparation used AI assistance; the submitting
contributor reviewed the changed lines and owns the behavior validation.