Avoid network access during ordinary tests

Nobody has claimed this yet.

## Proposed refactor

Some of our tests use the `MNIST` and `TrialMNIST` classes which download the MNIST data off the internet (or from a cache).

https://github.com/PyTorchLightning/pytorch-lightning/blob/948cfd24de4f64a2980395581f15544e5e37eab0/tests/helpers/datasets.py#L25

https://github.com/PyTorchLightning/pytorch-lightning/blob/948cfd24de4f64a2980395581f15544e5e37eab0/tests/helpers/datasets.py#L134

### Motivation

We should avoid this to reduce the inherent flakiness of network access.

### Pitch

If a test uses these classes, do one of 3 options:
- Remove the test, if we consider it's not useful (probably an old test)
- Update the test to use either `RandomDataset` or a `RandomMNIST` class which would "mock" the actual MNIST dataset but with random data.
- Move the test to `tests/benchmarks`

### Additional context

Master is currently blocked due to errors while reading the MNIST data zipfiles.

______________________________________________________________________

#### 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 @akihironitta

Start in tests/helpers/datasets.py at the MNIST and TrialMNIST definitions, then find ordinary tests that use them. Review each usage and decide whether it should be removed, use RandomDataset or a random MNIST substitute, or move to tests/benchmarks. Done means ordinary tests no longer access the network or depend on MNIST downloads.

Written by the indexing model from the issue text.