Support re-instantiation for custom DataLoader in Lightning

[awaelchli]

What does this PR do?

Part of #10329

Adds support for re-instantiating custom DataLoaders in Lightning, a feature that was recently added to Lite first in #10279.

• The utilities in Lite were moved to general data utilities
• The place in Lightning where we call *_dataloader on the LightningModule or DataModule gets wrapped with a context manager.

Limitations:
The patching will only work if the DataLoader gets instantiated inside the *_dataloader methods.
The custom DataLoader needs to be compatible with the patching, i.e., it needs to allow Lightning to write the arguments to it's dict.

Before submitting

• Was this discussed/approved via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or internal minor changes/refactorings)

PR review

Anyone in the community is free to review the PR once the tests have passed.
Before you start reviewing make sure you have read Review guidelines. In short, see the following bullet-list:

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

Did you have fun?

I made sure I had fun coding 🙃

Part of #1 (it's a lie, this is just here to avoid noisy GitHub bot)

cc @Borda @justusschock @awaelchli @ninginthecloud

[tchaton]

LGTM !

[rusty1s]

@awaelchli This is really cool. This comes in handy with our custom data loaders in PyG. Thank you!

One issue I ran into is the use of "nested inheritance" of data loaders:

torch.utils.data.DataLoader
       |
MyBaseDataLoader
       |
MyFinalLoader

This seems to do some weird things in the _replace_dataloader_init_method context manager since we overwrite the __init__ function of both data loader classes. I can try to set up a reproducible example later.

[carmocca]

Hi @rusty1s! Glad that you find it useful, we actually worked on this with PyG as a motivation.

Happy to take a look once you define what those "weird things" are. You can find existing tests here if you are interested in poking at it yourself:

https://github.com/PyTorchLightning/pytorch-lightning/blob/1acb0dcbf4720b49b3de557d75f57a06c7276083/tests/utilities/test_data.py#L153-L176

Also, it would be better if you can open a new issue once you have a repro. Thank you!