Fix ValueError: Assign retain set to eval in unlearn mode

[ma-kjh]

Add logic to retain evaluation data during unlearning.

What does this PR do?

This PR fixes a ValueError that occurs when running the unlearning process with eval_strategy="epoch" or "steps".

Problem

Currently, in src/data/__init__.py, when mode="unlearn", the code merges the splits into a single train dataset and then pops (removes) the original splits (including retain) from the data dictionary.
As a result, when the Trainer tries to access eval_dataset (typically the retain set), it finds None, causing the following error:
ValueError: You have set args.eval_strategy to IntervalStrategy.EPOCH but you didn't pass an eval_dataset to Trainer.

Solution

I added logic to assign the retain split to data["eval"] before popping the splits. This ensures that the Trainer has access to evaluation data during the unlearning process.

Fixes # (issue)
N/A

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Have you gone through the contributions guide?
• Are your changes documented? Read documentation guidelines here.