Changes as follows:

Overhaul datasets

• Adds dataset base class with lots of checks
• Datasets inherit from base class, simplifying their implementations
• Adds ImageFolder and TensorDataset classes
• More unified dataset tests
• Much better dataset docs + guide
• PatchDataset returns tensors and not tuples
• Clean thorough example for defining datasets

Overhaul examples + quickstart

• Quickstart examples + one very simple 5-minute quickstart tutorial (closes Clear quickstart examples #571 )
• Organises examples: basics section is now indeed basics; add "Models" and "Physics" sections
• Defining physics -> better cleaner example
• Training your own model -> better cleaner example (to sit alongside Trainer compatibility with Denoiser #593 )
• Pretrained reconstructors Guide + Example
• Finetuning example (to clean up after RAM #524 merged)
• README and index now cleaner
• Custom ordering

Minor fixes

• Closes Should this be ZeroNoise()? #621
• Closes load_example docs #669
• Closes MoDL docs failing #672
• Closes British English -> American English #656
• Full-reference metrics can handle nans
• DPIR has optional denoiser param
• Spelling mistakes standart -> standard
• Tomography accepts tensor angles better
• physics = GaussianNoise -> physics = Denoising(GaussianNoise)
• some small utils loading funcs
• TensorList shape, all, any and isnan
• Docs code blocks on lightmode appear slightly less gray
• Add autogenerated "References" to every example (discussed in Slack)

Notes

• Quickstart requires the redirect (rather than inserting auto_examples/basics/demo_quickstart into index toctree otherwise the left sidebar doesn't show up in the quickstart example??

Checks to be done before submitting your PR

• python3 -m pytest deepinv/tests runs successfully.
• black . runs successfully.
• make html runs successfully (in the docs/ directory).
• Updated docstrings related to the changes (as applicable).
• Added an entry to the CHANGELOG.rst.

Todo

• Add quickstart tutorial notebook to colab and replace original colab link in readme/index
• Switch all temp models over to RAM after RAM #524 merged
• finetuning example has more modalities from pretrained example. finetuning example links out to custom physics + data at end + what's next?
• "New to DeepInverse?" auto add to all examples?