Add label smoothing param in DiceCELoss

[dmtryzarubin]

Is your feature request related to a problem? Please describe.
It would be nice to have such ability to use label smoothing in init method of DiceCELoss

Describe the solution you'd like
Add label_smoothing param into DiceCELoss and pass it directly into torch.nn.CrossEntropyLoss

Describe alternatives you've considered
_

Additional context
This is the only way to do this now:

from monai.losses import DiceCELoss


criterion = DiceCELoss()
criterion.cross_entropy.label_smoothing = 0.1