Fix float casting when autoscaling int values to byte

[melissalinkert]

Fixes #4210.

Test code that uses the autoscale method:

$ cat AutoscaleTest.java 
import java.awt.image.BufferedImage;
import loci.formats.gui.AWTImageTools;

public class AutoscaleTest {
  public static void main(String[] args) throws Exception {
    int[] inputPixels = new int[] {40};
    BufferedImage input = AWTImageTools.makeImage(inputPixels, 1, 1, false);
    BufferedImage scaled = AWTImageTools.autoscale(input, 20, 120);
    byte[][] scaledPixels = (byte[][]) AWTImageTools.getPixels(scaled);
    System.out.println(scaledPixels[0][0]);
  }
}

Without this change, the value printed is 0. With this change, the value printed is 51.

The cast is the same strategy as used in the short[] case in line 1638.