Add --dry-run flag to `parler cleanup`

[AbdelStark]

Summary

Add a --dry-run flag to parler cleanup that shows what would be deleted without actually removing anything. This is a safe, confidence-building option for operators running cleanup in production environments.

Background

parler cleanup currently prunes stale run bundles and temp audio files immediately when invoked. There is no way to preview what will be removed before committing. A --dry-run mode is a standard UX pattern for any destructive CLI command and makes the tool easier to trust.

What to implement

Add a --dry-run flag to the cleanup command in parler/cli.py:

# Preview only — prints what would be removed, removes nothing
parler cleanup --dry-run --older-than-days 7

# Current behaviour unchanged
parler cleanup --older-than-days 7

Expected dry-run output (text mode):

[dry-run] Would remove 3 run bundle(s) and 1 temp audio file(s) older than 7.0 day(s).

Expected dry-run output (--json mode):

{
  "dry_run": true,
  "older_than_days": 7.0,
  "would_remove_runs": 3,
  "would_remove_temp_audio": 1
}

When --dry-run is set, the function should call the existing prune_* helpers with a counting approach but skip deletion — or add a dry_run=True parameter to prune_run_summaries and prune_managed_audio_files in parler/runlog.py and parler/audio/ingester.py.

Files to look at

File	What to understand
parler/cli.py — cleanup command (~line 956)	Where to add the --dry-run option and conditional logic
parler/runlog.py — prune_run_summaries()	The existing prune logic to make dry-run-aware
parler/audio/ingester.py — prune_managed_audio_files()	Same, for temp audio
tests/unit/test_cli_commands.py — TestOperationalCommands	Where to add tests

Acceptance criteria

• parler cleanup --dry-run prints what would be removed and exits 0
• Nothing is deleted when --dry-run is set
• --dry-run composes correctly with --older-than-days, --runs, --no-runs, --temp-audio, --no-temp-audio
• --dry-run --json emits a JSON object with a "dry_run": true key alongside the counts
• At least two unit tests covering dry-run behaviour (one text, one JSON)
• ruff check and ruff format --check pass

Setting up

git clone https://github.com/AbdelStark/mistral-parler && cd mistral-parler
uv sync --locked --group dev
uv run parler doctor          # verify environment
uv run pytest tests/unit/test_cli_commands.py -q   # confirm baseline

Tips for first-timers

• All CLI commands use Click. Adding a boolean flag is a one-liner: @click.option("--dry-run", is_flag=True, help="...")
• The cleanup command already receives runs: bool and temp_audio: bool flags — pattern your dry-run logic similarly
• Use CliRunner from click.testing for tests (see existing test_runs_list_show_and_cleanup for a reference)