Rework model docstrings for progressive disclosure of complexity

[mattdangerw]

We need to simplify our model docstring so they are more easily understandable, and we can start by porting the changes in #843 to all other models.

Let's do this one model at a time, to keep things granular.

Checklist:

• Make sure to update any "custom vocabulary" examples to match the model actual vocabulary type and special token requirements (varies per model).
• Test out all docstring snippets!
• Make sure to follow our code style guidelines re indentation etc.

[mattdangerw]

Also, we are getting a lot of contributor interest right now! Let's all try to..

• Use this issue to communicate.
• Only work on a single model at a time. E.g. wait till you have a PR ready for one model before starting on the next.
• Be patient with each other as we all self organize here :)

Thanks!

[kanpuriyanawab]

I can take this up for Roberta. Thanks.

[ADITYADAS1999]

Thanks for this issues & guideline, I want to work with f_net model.

[Cyber-Machine]

@mattdangerw I'd like to work with Distil_bert model. Thanks

[soma2000-lang]

@mattdangerw I am working on albert

[TheAthleticCoder]

I shall take this up for deberta_v3 then

[abuelnasr0]

I am working with GPT2

[Warlord-K]

@mattdangerw I'd like to take up opt.

[abuelnasr0]

I will take XLMRoberta instead of GPT-2 for now as it is not exported to keras_nlp library to be used in colab to test docsting snippets. I will continue my work with GPT-2 after installing dependencies in my local machine soon.