Add Sigmoid and Logit Transformers

[amontanez24]

Problem Description

The sigmoid and logit transformations are use in SDV to enforce the between constraint. They could also be used in RDT's NumericalTransformer to keep values within the bounds. It would be convenient to pull this code out into it's own transformer.

Expected behavior

For Logit something like

class LogitTransformer:
    def __init__(self, high, low):
        pass

    def transform(data):
        data = (data - low) / (high - low)
        data = data * 0.95 + 0.025
        data = np.log(data / (1.0 - data))
        return data

    def reverse_transform(data):
        data = 1 / (1 + np.exp(-data))
        data = (data - 0.025) / 0.95
        data = data * (high - low) + low
        data = data.clip(low, high)

The sigmoid would just swap the transform and reverse transform functions

[csala]

I think the proposal is right, but I have doubts about the 0.95 + 0.025 part, as well as the normalization performed before that.

At most, these could be set as optional arguments, but I would expect a LogitTransformer to just do that: apply a logit function, and then apply a sigmoid when reversing, without anything else.