This benchmark performs no work: it simply calls bench_start and bench_end back to back. This
shows us the overhead involved with the infrastructure, a combination of, e.g.:
- the overhead of calling an imported Wasm function
- the overhead of our measurement mechanism