Deepspeed ZERO MiCS support

Nobody has claimed this yet.

### Description & Motivation

After deepspeed 0.9.2, they provided [Misc support](https://deepspeed.readthedocs.io/en/latest/zero3.html#example-mics-configurations), which can specify how to split parameters across devices [reference](https://github.com/microsoft/DeepSpeed/discussions/3438). It's a pleasure if lightning deepspeed strategy has such features.

### Pitch

By detecting if the keywords 'mics_shard_size' or 'mics_hierarchical_params_gather' in deepspeed args, deepspeed can decide whether to use mics.
This requires using 'deepspeed.zero.MiCS_Init()' instead 'deepspeed.zero.Init()' when initializing the model.

### Alternatives

_No response_

### Additional context

I simply override DeepSpeedStrategy.model_sharded_context (line 519 in lightning.pytorch.strategies.deepspeed.py) with
, it works. So, can we add some conditions here to use an alternative context manager?

cc @borda @awaelchli

Start in lightning.pytorch.strategies.deepspeed.py at DeepSpeedStrategy.model_sharded_context around line 519, and review the referenced DeepSpeed MiCS configuration and initialization APIs. Verify how the existing strategy detects its arguments, then confirm that MiCS settings select the alternative context manager while other configurations retain current behavior.

Written by the indexing model from the issue text.