Update simulation examples to use flwr_datasets

[jafermarq]

This PR updates the simulation-pytorch and simulation-tensorflow examples. They now use the Flower datasets API. The examples run just fine. Some things to iron out before merging:

• Update URLs once datasets docs are live
• Update pyproject.toml once flwr_datasets is in PyPI
• Can we disable tqdm ?
• Reflect this PR in change log.
• Once happy with both TF and Pytorch sim.py --> apply changes to notebook

[adam-narozniak]

The jupyter notebooks files need to be updated.

For the tensorflow, I think it's better if we use this approach (instead of directly to tensors)

tf_dataset = partition.to_tf_dataset(columns="img", label_cols="label", batch_size=64,
shuffle=True) # model.fit has shuffle = True so that is also not 100% needed

Or alternatively .with_format("numpy").

One thing about the .with_format("tf") I'm not sure is what happens if the GPU is present (tf by default uses GPU if it's available, so I'm not sure if the full data goes to GPU when e.g. trainset['images'] because it fetches all the images and it happens later in the code)

[adam-narozniak]

LGTM! I like the change in the TF example, it looks clean.