Improve list of benchmarks

The benchmarks are currently describe as a list. We should put together a more comprehensive and detailed view.
A table, or a more complex visualization in which, for each benchmark, we list multiple features (like CL scenario in which it can be used, type of data, possible applications and so on).

[This is an example](https://robustbench.github.io/). The main problem there is to maintain updated the best accuracy. Due to the different CL scenarios available it would require a lot of effort.