[Feature Request] Add value_range parameter to some of the policy operations for better policy control

[iumyx2612]

I trained autoalbument with 11 epochs on my dataset and I think that's enough (the Average Parameter change oscillating with small amount). The output config has an augmentation with really big scale limit (around 9) and it's unusable on my dataset. So I added value_range to Scale operation like this:

class Scale(Operation):
    def __init__(self, temperature, value_range=(0 + 1e-8, 10.0)):
        super().__init__(temperature, value_range=value_range, is_spatial_level=True)

    def apply_operation(self, input, value):
        return F.scale(input, scale=value)

    def as_transform(self, value, p):
        return A.ShiftScaleRotate(
            shift_limit_x=(0, 0),
            shift_limit_y=(0, 0),
            rotate_limit=(0, 0),
            scale_limit=(value, value),
            p=p,
        )

so I can go in search.yaml to change the value of value_range