Centralize location of rank_zero utility functions

[ananthsub]

Proposed refactor

The rank_zero_* utilities are currently split between distributed.py and warnings.py
distributed.py
https://github.com/PyTorchLightning/pytorch-lightning/blob/8c07d8bf905e395bbd2142b5df7b185b8e936c41/pytorch_lightning/utilities/distributed.py#L47-L95

https://github.com/PyTorchLightning/pytorch-lightning/blob/8c07d8bf905e395bbd2142b5df7b185b8e936c41/pytorch_lightning/utilities/warnings.py#L23-L51

Proposal: Move all of the rank_zero_only utilities into their own module

Motivation

It is currently impossible to deprecate something from within distributed.py using the standard Lightning conventions without creating a circular import.

Creating a separate module solves this and clarifies dependencies better.

I'm facing this issue in https://github.com/PyTorchLightning/pytorch-lightning/pull/11745/files

Pitch

Create a new module: utilities/rank_zero.py which will house all of these decorators.

Alternatives

Move rank_zero_only utilities from warnings.py back into distributed.py

---

If you enjoy Lightning, check out our other projects! ⚡

• Metrics: Machine learning metrics for distributed, scalable PyTorch applications.

• Lite: enables pure PyTorch users to scale their existing code on any kind of device while retaining full control over their own loops and optimization logic.

• Flash: The fastest way to get a Lightning baseline! A collection of tasks for fast prototyping, baselining, fine-tuning, and solving problems with deep learning.

• Bolts: Pretrained SOTA Deep Learning models, callbacks, and more for research and production with PyTorch Lightning and PyTorch.

• Lightning Transformers: Flexible interface for high-performance research using SOTA Transformers leveraging Pytorch Lightning, Transformers, and Hydra.

cc @justusschock @awaelchli @akihironitta @rohitgr7