Improve Fraud Dataset

- The Fraud Dataset doesn't work well when all the rows and columns are used. The top 2 models generated have only 1 feature. The rest of the models have 66 or 12 features
- I realize this is a synthetic dataset so this may be intended (or not a problem). Feel free to close this issue if that is the case.

## With 2500 Rows, Some columns (following docs example)
```python
import evalml
from evalml import AutoClassificationSearch
from evalml.objectives import FraudCost

fraud_objective = FraudCost(retry_percentage=.5,
interchange_fee=.02,
fraud_payout_percentage=.75,
amount_col='amount')
X, y = evalml.demos.load_fraud(n_rows=2500)
X = X.drop(['datetime', 'expiration_date', 'country', 'region', 'provider'], axis=1)
X_train, X_holdout, y_train, y_holdout = evalml.preprocessing.split_data(X, y, test_size=0.2, random_state=0)
automl = AutoClassificationSearch(objective=fraud_objective,
additional_objectives=['auc', 'recall', 'precision'],
max_pipelines=5)

automl.search(X_train, y_train)
automl.describe_pipeline(automl.rankings.iloc[0]["id"])
```

## With All Rows, All columns
```python
X, y = evalml.demos.load_fraud()
X_train, X_holdout, y_train, y_holdout = evalml.preprocessing.split_data(X, y, test_size=0.2, random_state=0)
automl = AutoClassificationSearch(objective=fraud_objective,
additional_objectives=['auc', 'recall', 'precision'],
max_pipelines=5)
automl.search(X_train, y_train)
automl.describe_pipeline(automl.rankings.iloc[0]["id"])
```