Update content of task pages (#543)
* Misc changes for fill mask
- Update to distilroberta-base
- Change schema to a more intuitive example
- Removed too many digits in probabilities
- Updated short paragraph to avoid diving too much into details of training (which is not too important for our audience)
- The about.md was assuming knowledge in things such as "fine-tuning", "pre-training", etc. I tried to water it down.
- Added some emojis and changed some wording to make it a bit more fun
* Misc updates for question answering
- Updated short sentence at top as I don't think saying QA is a NLU task is super meaningful.
- I'm not super happy with my proposal though, recommendations are welcomed
- Simplified text from models (BERT, SQUAD, etc are not known by the audience)
- Used QA instead of Question ANswering many times to avoid repeating it too many times
- Restructured Task Variants as it was a bit hard to read (used bullet points)
- Should we consider having a table instead?
- Remove assumption people know about transformers library
- Added links to other tasks
