Benchmark baseline definition, allowed deviation and continous validation

[tobrun]

Thanks to @baparham, we now have a MVP benchmark implementation from #51. This approach uses the benchmark package so that executing dart run benchmark runs all the *_benchmark.dart within the benchmark directory. Having this setup is a great as a development tool, it allows us to quantify performance differences between different implementations. Iterating on this further, we can invest into making this a validation system that runs in CI.

Work to be done:

• integrate running benchmark in CI env. + store results as CI run artifact
• run integration, create a baseline file from that contains benchmark name, output value and allowed deviation percentage
• integrate validation step that verifies that output didn't breach the baseline deviated output value, fail run if it did

cc @lukas-h