Add a microbenchmark for type inference / dataflow

[msridhar]

This will help us to identify performance issues in our dataflow analysis that could improve NullAway's performance on large methods. The benchmark was generated with a script.

[lazaroclapp]

The benchmark was generated with a script.

If so, do we want to commit the generated code, or add the script itself to our build process and have it add a resources/generated/... directory with this sort of generated test code?

[msridhar]

If so, do we want to commit the generated code, or add the script itself to our build process and have it add a resources/generated/... directory with this sort of generated test code?

So the script was a one-off hack and was not meant to be generalizable. I don't foresee us doing a lot of tweaking to this benchmark. It's possible that we would want more micro-benchmarks like this one in the future. Maybe we can say that when we want to add a second benchmark, we will bite the bullet and check in the generator scripts and run them during the build? To me, that seems like overkill at this point, and I'd rather just commit the generated code for now.