Add is_initialized method and refactor

[muellerzr]

Add a static is_initialized method and refactor existing logic for initialization check

What does this add?

This PR introduces a static is_initialized method which will check if the state's dictionary is empty or not. It also refactors current methods to have initialized be a @property check of the same to increase code clarity towards what is happening under the hood.

Who is it for?

Closes #933

Why is it needed?

The feature request asked for a similar API to what torch uses for cuda, i.e. torch.cuda.is_initialized() for Accelerate, so a user can do import accelerate; accelerate.state.is_initialized()

What parts of the API does this impact?

User-facing:

A new is_initialized method was added to the accelerate.state module

Internal structure:

initialized was changed to be a dictionary check to mimic how the new is_initialized() function operates

Basic Usage Example(s):

# One way:
import accelerate
accelerate.state.is_initialized()

# Or:
from accelerate import Accelerator
accelerator = Accelerator()
accelerator.state.initialized

When would I use it, and when wouldn't I?

When a static method is needed to check if the AcceleratorState has been initialized without doing a getattr, or a cleaner interface to present to the user

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[sgugger]

Thanks for adding this!