GreaterThan constraint's transform strategy fails on columns that are not float

[amontanez24]

Environment Details

Please indicate the following details about the environment in which you found the bug:

• SDV version: 0.10.0
• Python version: 3.x
• Operating System: Windows, MacOS

Error Description

Related to #421

When a GreaterThan constraint is provided transform as the handling_strategy, it will fail if the high column is not a float. This is because the HyperTransformer reverse transforms the columns before the constraint's reverse transformation takes place. If the high column was initially an int or datetime, it gets transformed by the HyperTransformer. The constraint then has a value in the wrong form when attempting to apply its reverse transformation.

Steps to reproduce

from sdv.demo import load_tabular_demo
from sdv.constraints import UniqueCombinations
from sdv.tabular import GaussianCopula
from sdv.constraints import GreaterThan

employees = load_tabular_demo()

age_gt_age_when_joined_constraint = GreaterThan(
    low='age_when_joined',
    high='age',
    handling_strategy='transform'
)
constraints = [
    age_gt_age_when_joined_constraint
]

gc = GaussianCopula(constraints=constraints)
gc.fit(employees)

sampled = gc.sample(10, conditions={'age': 40})