string.drop_start behaves wrong on the JS target

```gleam
import gleam/io
import gleam/string

pub fn main() {
io.println(string.drop_start("广州abcdefghijklmn", 0))
io.println(string.drop_start("广州abcdefghijklmn", 1))
io.println(string.drop_start("广州abcdefghijklmn", 2))
io.println(string.drop_start("广州abcdefghijklmn", 3))
}
```

outputs on the JS target:

```
广州abcdefghijklmn
bcdefghijklmn
efghijklmn
fghijklmn
```

So the first two characters are counted as 3 each. `unsafe_byte_slice` is used here with _byte_ offsets: https://github.com/gleam-lang/stdlib/blob/c6f7ceee9c569c71849d7e4c5a94e535feaae724/src/gleam/string.gleam#L236

It calls `string_byte_slice`, which contrary to its name does not operate on bytes but UTF-16 _code units_: https://github.com/gleam-lang/stdlib/blob/c6f7ceee9c569c71849d7e4c5a94e535feaae724/src/gleam_stdlib.mjs#L202

Thus the wrong offsets are sliced from the string.