Regression in round-tripping UTF-16 strings after JDK-8311906

I'm seeing a regression when writing a string to a file as UTF-16 and reading it back in using `Files.readString(String, Charset)` after JDK-8311906 (Improve robustness of String constructors with mutable array inputs).

Repro:

```
import static java.nio.charset.StandardCharsets.UTF_16;

import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.util.Arrays;

public class Y {
  public static void main(String[] args) throws IOException {
    String original = "🤸";
    Path tmp = Files.createTempFile("tmp", ".txt");
    Files.writeString(tmp, original, UTF_16);
    String actual = Files.readString(tmp, UTF_16);
    System.err.printf("expected (%s), was (%s)\n", original, actual);
    System.err.printf("expected UTF_16 bytes: %s\n", Arrays.toString(original.getBytes(UTF_16)));
    System.err.printf("actual UTF_16 bytes: %s\n", Arrays.toString(actual.getBytes(UTF_16)));
  }
}
```

$ java -fullversion
openjdk full version "21.0.2+13-58"
$ java Y
expected (🤸), was (🤸)
expected UTF_16 bytes: [-2, -1, -40, 62, -35, 56]
actual UTF_16 bytes: [-2, -1, -40, 62, -35, 56]

$ java -fullversion
openjdk full version "22+35-2369"
$ java Y
expected (🤸), was (>Ø8Ý)
expected UTF_16 bytes: [-2, -1, -40, 62, -35, 56]
actual UTF_16 bytes: [-2, -1, 0, 62, 0, -40, 0, 56, 0, -35]