Fixed save_image to correctly output values in the full byte range.

[bobbens]

After changing this I saw the issue #24 which mentions this and mentions adding a range option. For now I've just removed the .mul(0.5).add(0.5) as without it, the follow code does not properly save the image.

import torch
import torchvision
from PIL import Image
I  = Image.open( './lena.jpg' )
I2 = torchvision.transforms.ToTensor()( I )
torchvision.utils.save_image( I2, 'test.jpg' )