CI: Cleanup benchmark CI scripts

[Robbepop]

#422 added support for wasmi CI benchmarking.

While the whole thing seems to be working fine and outputs well formatted markdown code that is pretty readable the scripts that extract the benchmark information out of Criterion output are a bit unreadable.
Also those scripts depend on instable Criterion console output instead of depending on the more stable machine readable output documented here:
https://bheisler.github.io/criterion.rs/book/cargo_criterion/external_tools.html

We should make Criterion output JSON that our scripts could consume with tools such as jq (JSON query) in order to create the pretty printed markdown table.

This has 2 advantages:

• Making current scripts more readable.
• (More important) Make the CI less brittle and instable to changes in Criterion console output.

[Robbepop]

Closed since #464 got merged.