perf: introduce benchmarks

[jasonkuhrt]

This contributes long-term toward tracking and improving issues like what #310 appeared to be at first (it later turned out to be a bug manifesting somewhat like bad perf, in that case).

TODO

• consider removing path relative util functions as their help may not outweigh their surprise
• optionally save benchmark results to JSON data that can be easily charted somewhere
• setup ci to run and commit benchmarks for every master commit
• add architecture docs about how to work on the benchmark system
• add reference docs about how to add new benchmarks
• flow benchmarks
• final sanity check on diff, refactor/clean as needed

[jasonkuhrt]

One of these or other tools might be useful to visualize benchmark results over time:

• https://www.taucharts.com/
• https://www.chartjs.org/

We could create a sandbox at https://codesandbox.io/ to host the charts. Maybe pull down the JSON from github raw link.

[jasonkuhrt]

Not going to do the the ci automation part yet, because:

• its a monorepo, not all changes will be related to graphqlgen

• enforcing that every second commit become a benchmark history update might be really annoying for the git history... maybe it should only be done for e.g. perf-type and feature-type commits

• maybe we should use CI to generate a PR comment containing benchmark runs, that comment being updated upon every PR commit...

I think these things need more discussion. For now we can start with the manual process.

--edit

issue made #375