Refactor dataloading

[ethanwharris]

Before submitting

• Was this discussed/approved via a Github issue? (no need for typos, doc improvements)
• Did you read the contributor guideline?
• Did you make sure to update the docs?
• Did you write any new necessary tests?
• If you made a notable change (that affects users), did you update the CHANGELOG?

What does this PR do?

Fixes #953 Fixes #840 Fixes #698

• Refactored calls to reset_train_dataloader and reset_val_dataloader to only happen when needed
• Slightly changes the logic for disable validation to ignore the validation_dataloader length
• Removed default unpacking of dataloader and adding of RandomSampler - see comment in refactor len(datasets) call. #953
• Changed Exception order so that MisconfigurationException will be called and given when using IterableDataset
• Removed dependency on Dataloader.dataset following Handle abstract loader that doesn't have a dataset member #840
• num_training_batches =float('inf') is now the default when train dataloader doesn't have __len__ (in addition to when using an IterableDataset)
• Added test for training using an iterable

PR review

Anyone in the community is free to review the PR once the tests have passed.
If we didn't discuss your PR in Github issues there's a high chance it will not be merged.

Did you have fun?

Make sure you had fun coding 🙃

[Darktex]

Does this need to be considered infinite? Normally, IterableDatasets are finite, we just don't know how long they are the first epoch. Another way of doing this would be to set it to infinite (or -1, or whatever placeholder value works best) and keep counting how many steps we did the first epoch. Once we start the second epoch, we can print a bar and all since we know the length will not change

[ethanwharris]

Yeah, thinking about it that should totally just be has_len or similar. No problem with the idea of keeping a record of the number of steps - although I would probably opt for that to be done in a seperate PR (lile when we add IterableDataset support for val and test) - but I'm happy to add that now if desired

[williamFalcon]

@ethanwharris nice job.

I'll merge this, then fix the TPU stuff then you can add what @Darktex suggested!
Running GPU tests now