Add output argument to Trainer.predict and remove DataPipelineState

[ethanwharris]

What does this PR do?

• Adds output argument to Trainer.predict
• Adds collate_fn and input_transform properties to the DatasetProcessor base class
• Removes ProcessState (refacors text data to apply tokenization in the collate function instead)
• Removes DataPipelineState
• Trims DataPipeline

With the state gone, there is now no more magic happening in the background. This means that properties such as labels and parameters are required to be passed to the task by the user.

Fixes #920

Before submitting

• Was this discussed/approved via a Github issue? (no need for typos and docs improvements)
• 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?
• Did you write any new necessary tests? [not needed for typos/docs]
• Did you verify new and existing tests pass locally with your changes?
• If you made a notable change (that affects users), did you update the CHANGELOG?

PR review

• Is this pull request ready for review? (if not, please submit in draft mode)

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 🙃

[codecov]

Codecov Report

Merging #1157 (055cf7d) into master (defbace) will decrease coverage by 0.29%.
The diff coverage is 94.73%.

❗ Current head 055cf7d differs from pull request most recent head 71773c7. Consider uploading reports for the commit 71773c7 to get more accurate results

@@            Coverage Diff             @@
##           master    #1157      +/-   ##
==========================================
- Coverage   89.26%   88.96%   -0.30%     
==========================================
  Files         286      285       -1     
  Lines       13045    12745     -300     
==========================================
- Hits        11644    11339     -305     
- Misses       1401     1406       +5     

Flag	Coverage Δ
unittests	88.96% <94.73%> (-0.30%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
flash/audio/classification/cli.py	100.00% <ø> (ø)
flash/audio/speech_recognition/data.py	100.00% <ø> (ø)
flash/core/data/process.py	64.00% <ø> (+1.93%)	⬆️
flash/core/data/splits.py	96.66% <ø> (-0.48%)	⬇️
flash/core/integrations/icevision/data.py	95.55% <ø> (-0.19%)	⬇️
flash/core/regression.py	86.36% <ø> (-0.60%)	⬇️
flash/core/utilities/stages.py	70.00% <ø> (-3.92%)	⬇️
flash/graph/classification/cli.py	92.30% <ø> (ø)
flash/image/classification/adapters.py	82.62% <ø> (ø)
flash/image/classification/cli.py	84.61% <ø> (ø)
... and 99 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update defbace...71773c7. Read the comment docs.