feat(model): Implement safe model deletion with trash functionality

[SyedaAnshrahGillani]

This pull request introduces a new "trash" feature to the ART framework,
providing a safety net for model deletion. Previously, deleting a model was
a destructive and irreversible operation. Now, models are moved to a .trash
directory, allowing for recovery or permanent deletion at a later time.

This PR includes the following changes:

• Safe Deletion: The delete_model function in src/art/model.py now moves
  models to a .trash directory instead of permanently deleting them.
• Trash Management: New functions have been added to src/art/model.py to
  manage the trash:
  • list_trash: Lists all models in the trash.
  • restore_from_trash: Restores a model from the trash.
  • empty_trash: Permanently deletes all models from the trash.
• CLI Integration: The new trash functionality is exposed through the CLI in
  src/art/cli.py with the following commands:
  • delete: Moves a model to the trash.
  • list-trashed-models: Lists all models in the trash.
  • restore-model: Restores a model from the trash.
  • empty-trashed-models: Permanently deletes all models from the trash.

This feature significantly improves the safety and usability of the ART
framework by preventing accidental data loss.

[Mirza-Samad-Ahmed-Baig]

Good Catches.

[arcticfly]

Nice! How does this apply to the model.delete_checkpoints method? I can imagine a backup of old checkpoints that isn't automatically synced to s3 being useful.

[SyedaAnshrahGillani]

Nice! How does this apply to the model.delete_checkpoints method? I can imagine a backup of old checkpoints that isn't automatically synced to s3 being useful.

That's a great question,

You're right, the current implementation of the trash feature is focused on
the model as a whole and doesn't extend to the delete_checkpoints method.
The delete_checkpoints function still deletes the checkpoints permanently
from the local filesystem.

I think applying a similar "trash" concept to checkpoints is an excellent
idea for a future enhancement. It would provide a safety net for local
checkpoints that haven't been synced to S3. I can create a new issue to
track this feature.