Unroll dict input before call Accelerator X_steps and update function type

[four4fish]

Proposed refactor

From @justusschock's comments in #10890
Refactor accelerator X_steps()'s signature, to support both positional and keyword arguments
Instead of unrolling step_kwarg dictionary in accelerator steps, unroll them in caller side.

This will unblock #10648 move functions from accelerator to strategies part

Motivation

Keep flexibilities and improving function typing

Pitch

Current Accelerator X_step typing signature is
https://github.com/four4fish/pytorch-lightning/blob/master/pytorch_lightning/accelerators/accelerator.py#L124-L148

Update to :
def X_step(self, *args, **kwargs) -> Optional[STEP_OUTPUT]:
Which not align with what torch model training step and model() signature.

In caller side, eg: optimizer_loop.py. Instead of passing dictionary step_kwargs, pass *step_kwargs.values()

Additional context

---

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