Enhancements to make this more valuable to GraphQL backend developers

[coco98]

• Work your way up to a 95% percentile latency limit to automatically find the rps where the system breaks
• Sample CPU and RAM (even if inaccurate) to give the developer a sense of resource consumption
• Spin up and spin down a backend for every suite so that resources are freed up

[0x777]

1. I agree.

2. We can't do this. There is no guarantee that the server that is being bench-marked is running on the same machine as graphql-bench.

3. We should have pre/post benchmarking scripts that can be executed

[remorses]

2. You can take cpu and ram info from docker, there is a python docker api that can be used to get the container used resources