[RFC] "auto" precision support

Nobody has claimed this yet.

## 🚀 Feature
Introducing the support for `precision="auto"`, which enables AMP whenever available and falls back to fp32 otherwise.

### Motivation

After introducing `accelerator="auto"` support, it is hard to determine the precision type to use.
Personally, I use AMP, whenever running on GPU and the GPU supports it. Otherwise I use fp32. Since precision has to be set at the same time, we specify the accelerator, with `accelerator="auto"` we cannot easily at that time determine on which accelerator it will be running (well we could, but that would be kind of duplicating logic we use for accelerator-detection internally) without a larger amount of boilerplate.

### Pitch

Have `precision="auto"` to switch between AMP and fp32 depending on the accelerator and whether the accelerator on hand does support AMP.

### Alternatives

Manually write out all the boilerplate.

______________________________________________________________________

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

- [**Metrics**](https://github.com/PyTorchLightning/metrics): Machine learning metrics for distributed, scalable PyTorch applications.

- [**Lite**](https://pytorch-lightning.readthedocs.io/en/latest/starter/lightning_lite.html): 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**](https://github.com/PyTorchLightning/lightning-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**](https://github.com/PyTorchLightning/lightning-bolts): Pretrained SOTA Deep Learning models, callbacks, and more for research and production with PyTorch Lightning and PyTorch.

- [**Lightning Transformers**](https://github.com/PyTorchLightning/lightning-transformers): Flexible interface for high-performance research using SOTA Transformers leveraging Pytorch Lightning, Transformers, and Hydra.

cc @borda @tchaton @justusschock @awaelchli @carmocca @akihironitta @rohitgr7 @kaushikb11 @ananthsub

Start by tracing the existing accelerator="auto" support and the code that handles AMP precision selection. Determine where precision is resolved relative to accelerator detection, then verify that precision="auto" selects AMP when supported and fp32 otherwise with suitable tests.

Written by the indexing model from the issue text.