[BMG]: Please add a guidelines to perform fine-tuning and inferencing on multi-gpu BMG machines

[raj-ritu17]

Describe the bug A new guidelines is needed to perform fine-tuning and inferencing on multi-gpu BMG machines.

Issue:

1. Now, we are using following xpu libraries -->[bmg-xpu]( pip install --pre --upgrade ipex-llm[xpu_2.6] --extra-index-url https://download.pytorch.org/whl/xpu)
2. which doesn't requires additional one-api installation and other packages.
3. most of the document related to gpu parallelism to perform fine-tune and inference are based on old xpu version
4. such as -->example
5. inf-mutli-gpu
6. finetuning-multi-gpu-bmg
7. need an additional documentation for multi-gpu support for BMG machines.