Fix import for XCCl availability check

[D4idalos]

fix: correct ImportError in train_utils by updating accelerate.utils imports

This PR fixes a blocking ImportError in llama_cookbook/utils/train_utils.py that occurs when attempting to initialize training (specifically tested with QLoRA). The script tries to import is_ccl_available from accelerate.utils, but in current environments/versions of accelerate, this attribute is named is_xccl_available.

Motivation and Context:
While running fine-tuning on XPU-equipped clusters (like EuroHPC), the training process crashes immediately during the import phase. Correcting this import allows the train_utils.py to correctly identify the availability of the Collective Communications Library (CCL) for XPU backends.

Error Log:

File ".../llama_cookbook/utils/train_utils.py", line 26, in <module>
    from accelerate.utils import is_xpu_available, is_ccl_available
ImportError: cannot import name 'is_ccl_available' from 'accelerate.utils'
```

Feature/Issue validation/testing
I have verified the fix by running a QLoRA fine-tuning session.

[x] Test A: QLoRA Fine-tuning Initialization

Model: `meta-llama/CodeLlama-7b-Instruct-hf`

Environment: Python 3.11, requirements  installed.

Result: Before the fix, the script failed with ImportError. After changing the import to is_xccl_available, the training pipeline initializes correctly and the model starts training as expected.

Before submitting
[x] This PR fixes a typo or improves the docs.

[] Did you read the [contributor guideline](https://github.com/meta-llama/llama-cookbook/blob/main/CONTRIBUTING.md), Pull Request section?

[ ] Was this discussed/approved via a Github issue? Please add a link to it if that's the case.

[ ] Did you make sure to update the documentation with your changes?

[ ] Did you write any new necessary tests?