fast string

[bobzhang]

some benchmarks:

Estimated testing time 40s (4 benchmarks x 10s). Change using -quota SECS.
┌──────────────────────────────┬──────────┬─────
│ Name                         │ Time/Run │ Percentage │
├──────────────────────────────┼──────────┼─────
│ built-in compare             │  17.24ns │     96.49% │
│ length compare               │   3.83ns │     21.44% │
│ same_length built-in compare │  17.87ns │    100.00% │
│ same_length compare          │   5.91ns │     33.06% │
└──────────────────────────────┴──────────┴─────

From here we can see that OCaml built-in string comparison behaves worst when the length is same.
Our customized comparison is fast enough (3x times faster) to justify such stub inclusion, which will boost performance a lot in performance critical code

[bobzhang]

a more comprehensive benchmark, our local version will perform better when the length is long:

Estimated testing time 1.33333m (8 benchmarks x 10s). Change using -quota SECS.
┌───────────────────────────────────────────┬──────────┬────────────┐
│ Name                                      │ Time/Run │ Percentage │
├───────────────────────────────────────────┼──────────┼────────────┤
│ File "test.ml", line 19, characters 30-37 │  17.34ns │     98.56% │
│ File "test.ml", line 22, characters 30-37 │   6.01ns │     34.17% │
│ File "test.ml", line 25, characters 30-37 │   6.02ns │     34.23% │
│ File "test.ml", line 28, characters 30-37 │   3.56ns │     20.24% │
│ File "test.ml", line 32, characters 30-37 │  17.59ns │    100.00% │
│ File "test.ml", line 35, characters 30-37 │   6.38ns │     36.29% │
│ File "test.ml", line 38, characters 30-37 │   6.34ns │     36.02% │
│ File "test.ml", line 41, characters 30-37 │   6.05ns │     34.42% │

[bobzhang]

we change to a faster string comparison function in this commit, it may change the -bs-sort order (cc @chenglou )