Add a Divan adapter

[epompeii]

https://github.com/nvzqz/divan

For now the adapter will need to parse the text output: https://github.com/nvzqz/divan/issues/10

It would be nice to have https://github.com/bencherdev/bencher/issues/240 complete or once implemented do auto-tagging of the generic benchmarks. From the announcement post:

scratch        fastest  │ slowest  │ median   │ mean     │ samples │ iters
├─ from_str             │          │          │          │         │
│  ├─ &str     0.738 ns │ 0.799 ns │ 0.759 ns │ 0.757 ns │ 100     │ 409600
│  ╰─ String   26.8 ns  │ 32.18 ns │ 30.39 ns │ 30.42 ns │ 100     │ 25600
╰─ init_array           │          │          │          │         │
   ├─ 1000     572.5 ns │ 598.6 ns │ 583 ns   │ 584.5 ns │ 100     │ 800
   ├─ 2000     1.155 µs │ 1.197 µs │ 1.166 µs │ 1.165 µs │ 100     │ 400
   ╰─ 3000     1.759 µs │ 1.801 µs │ 1.77 µs  │ 1.77 µs  │ 100     │ 400

This would add tags &str and String for from_str; 1000, 2000, and 3000 for init_array.

Note though that each generic benchmark would be "monomorphized". There would be both from_str<&str> and a from_str<String> benchmark created in Bencher.

[nvzqz]

Heads up that I'm planning to get JSON output published soon. Parameters (args, types, thread count) will be properties of benchmark runs rather than part of the benchmark path. The last component of the benchmark path will be the benchmarked function. It will be up to programs consuming the output to determine how they want to present parameters.