Add fmt runtime benchmark

[Kobzol]

It would be really nice to have a benchmark that tests the performance of std::fmt. I haven't found any reasonable real-world benchmarks that I could extract yet (maybe something from rustdoc? but these codes tend to have a lot of dependencies), so I just included this very small "smoke test" for now. I expect that we will modify/replace it later, but at least we have something for now.

Finished fmt/fmt-debug-derive (1/2)
    [Instructions]: min:   4,903,829,976    mean:    4,903,829,980    stddev:           5
          [Cycles]: min:   1,314,690,611    mean:    1,328,393,111    stddev:  20,201,916
  [Wall time [µs]]: min:         322,033    mean:          324,207    stddev:       2,697
   [Branch misses]: min:       3,105,554    mean:        3,136,476    stddev:      16,651
    [Cache misses]: min:         488,771    mean:          605,165    stddev:      67,914

Finished fmt/fmt-write-str (2/2)
    [Instructions]: min:   3,408,031,576    mean:    3,408,031,578    stddev:           1
          [Cycles]: min:     858,696,891    mean:      863,965,390    stddev:   8,381,552
  [Wall time [µs]]: min:         233,006    mean:          235,397    stddev:       4,027
   [Branch misses]: min:          51,930    mean:           53,910    stddev:       2,234
    [Cache misses]: min:          28,441    mean:           29,505    stddev:       1,483