Registry Strategy with strategy base name

[four4fish]

Proposed refactor

Now we have {strategy_name}+{parameter} registries, but not {strategy_name} for most strategies. For example “ddp_find_unused_parameters_false” is registries but not “ddp”. The accelerator_connector uses both strategy_registry and enums to find str to strategy matching.
part of #10416

Motivation

Unify strategy registry behavior, remove enums and enable accelerator_connector logic simplification

Pitch

Step 1: define distributed_backend for every strategy. Like in DDPStrategy
https://github.com/PyTorchLightning/pytorch-lightning/blob/f41d1e5e5ebb7040a39d137695e818cada9a9234/pytorch_lightning/strategies/ddp.py#L83

Step 2: Registry strategy with the strategy.distributed_backend

    @classmethod
    def register_strategies(cls, strategy_registry: Dict) -> None:
        strategy_registry.register(
            cls.distributed_backend,
            cls,
            description=f"{cls.__class__.__name__} Strategy",
        )

Step 3: part of the #11449
Simplify accelerator_connector logic

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 @rohitgr7