BUG: bayesiansearch requires `Dataset.create_train_test` to have been called

[andersbogsnes]

Describe the bug

To Reproduce
Steps to reproduce the behavior:

from sklearn.ensemble import RandomForestClassifier
from ml_tooling.data import load_demo_dataset
from ml_tooling.search import Integer
from ml_tooling import Model

iris_data = load_demo_dataset("iris")

model = ml.Model(RandomForestClassifier())

best_model, results = model.bayesiansearch(iris_data, param_distributions={"max_depth": Integer(1, 200)})

Expected behavior
Should not raise exception

Additional context
iris_data.train_x is None, and thus it fails - we don't call score_estimator, so it's not automatic