The following contrived example shows a change in the behaviour of Random#ints after JDK-8248862.

The documentation for Random#ints and #longs specifies implementations for the methods:

https://github.com/openjdk/jdk/blob/09bfbf80639f059563fcd4432995b8c380cea298/src/java.base/share/classes/java/util/Random.java#L927-L942

https://github.com/openjdk/jdk/blob/09bfbf80639f059563fcd4432995b8c380cea298/src/java.base/share/classes/java/util/Random.java#L1043-L1063

The implementation of long is documented as having additional logic to 'reject over-represented candidates'. After the implementation of JDK-8248862, it looks like both methods are using logic in RandomSupport (in boundedNextInt and boundedNextLong) that are identical aside from the use of 'int' and 'long', and both of those implementations include the 'over-represented candidates' logic.

Is this change deliberate? If so, would it make sense to update the javadoc for Random#ints to clarify that, like #longs, it includes the additional logic?

```
import java.util.Random;

public class R {
  public static void main(String[] args) {
    Random r = new Random() {
      @Override public int nextInt() { return 9; }
      @Override public int nextInt(int i) { return nextInt(); }
      @Override public long nextLong() { return nextInt(); }
    };

    r.longs(400, 1000).limit(1).forEachOrdered(System.err::println);
    r.ints(400, 1000).limit(1).forEachOrdered(System.err::println);
  }
}
```

./jdk-17-ea+17/bin/java R
404
404

./jdk-17-ea+16/bin/java R
404
409 // this matches 'nextInt(n) + origin' in the Random#ints docs