Add AWS DataParellism

[nomDeZeus]

Hi!
I currently use AWS SageMaker to train my PL models. I recently found out this link :
https://aws.amazon.com/fr/blogs/aws/managed-data-parallelism-in-amazon-sagemaker-simplifies-training-on-large-datasets/
SageMaker provides its own implementation of DDP and I think that would be nice to be able to use it with PL :)
I looked into PL code and I think I could add this feature by extending pytorch_lightning.accelerators.accelerator.Accelerator. Does it seems like a good way to implement it? Are there some general advices/guidance you could give me about this?
If the you are interested in this feature, I can make a PR once I'm done with it.
Thank you!

Rémi

[carmocca]

In progress in #6271