[Feature] Support HF accelerate and DeepSpeed for inference

[ivan-khvostishkov]

Using HF accelerate or DeepSpeed engine for inference:

https://github.com/aws/amazon-sagemaker-examples/blob/main/inference/generativeai/llm-workshop/lab1-deploy-llm/intro_to_llm_deployment.ipynb

Also compare with this DeepSpeed example:
https://github.com/aws/amazon-sagemaker-examples/blob/main/advanced_functionality/pytorch_deploy_large_GPT_model/GPT-J-6B-model-parallel-inference-DJL.ipynb

See:

https://github.com/deepjavalibrary/djl-serving

[ivan-khvostishkov]

Released in v2.1.0:

https://github.com/aws-samples/sagemaker-ssh-helper/blob/v2.1.0/tests/test_frameworks.py#L764-L838
https://github.com/aws-samples/sagemaker-ssh-helper/blob/v2.1.0/tests/test_frameworks.py#L673-L709