Fix and simplify semantic-segmentation example

[qubvel]

What does this PR do?

It seems like the semantic-segmentation example is a bit outdated and doesn't run without fixing some issues. This PR does the following to improve the current state of the example:

1. Added instructions for accessing the sidewalk dataset to the semantic-segmentation example's README.md
2. Fixed missing dependencies required to run the semantic-segmentation example (accelerate, Pillow)
3. Fixed failing image transforms; replaced implemented transforms with those from albumentations
4. It was also observed that the "mean_iou" metric computation takes a lot of time and can be optimized. To improve this, I created a PR in the Evaluate project (Speeding up mean_iou metric computation evaluate#569). As soon as the fix is merged, the time to perform training will be reduced by half.

Updated scripts have been evaluated and the fine-tuned model achieved similar metrics to the previous one:

• previous model: https://huggingface.co/nielsr/segformer-finetuned-sidewalk-10k-steps
• new model: https://huggingface.co/qubvel-hf/segformer-finetuned-sidewalk-10k-steps

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

@amyeroberts @NielsRogge

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[amyeroberts]

Very nice! Thanks for working on this and improving our examples!