Call _handle_nullable_types in fit, transform, predict

[tamargrey]

The first step in using the component-specific nullable type handling will be to actually call _handle_nullable_types for the components that need it. This will involve updating the necessary components and any resulting test changes.

class MyTransformer(Transformer):
    # We can handle nullable types automatically if we want to
    def fit(self, X, y):
      ...
      X, y = self._handle_nullable_types(X, y)
      ...

# A component with more specific needs 
class MyEstimator(Estimator):
    # We can handle nullable types automatically if we want to 
    def predict(self, X):
      ...
      X, y = self._handle_nullable_types(X)
      ...

As part of this ticket we should also

• Understand the impact to automl search runtimes (expected to be increased, but the nullable types transformer would still be doing the actual transforming)
• Test use of nullable types at the pipeline level

This should not be merged into main until the rest of the needed changes to automl search to remove existing handlings
have been made. They should all be merged to main at the same time.