Add some AndroidX benchmark benchmarks #105
Description
A lot of our current assumptions for generated code are based on static analysis and not real-world runtime performance. Usually one is a good approximation for the other, but not always. We should empirically test our choices to ensure they are performing as expected.