Labels are sometimes cut off in ConfusionMatrix

Describe the bug When a dictionary of encoding-to-string mappings is provided to the ConfusionMatrix so that the class names will appear as labels, sometimes the class and axis labels get cut off.

To Reproduce

from sklearn.datasets import load_iris
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split as tts

from yellowbrick.classifier import ConfusionMatrix


iris = load_iris()
iris_X = iris.data
iris_y = iris.target
iris_classes = iris.target_names
i_X_train, i_X_test, i_y_train, i_y_test = tts(
    iris_X, iris_y, test_size=0.2
)
model = LogisticRegression()
iris_cm = ConfusionMatrix(
    model, classes=iris_classes,
    label_encoder={0: 'setosa', 1: 'versicolor', 2: 'virginica'}
)
iris_cm.fit(i_X_train, i_y_train)
iris_cm.score(i_X_test, i_y_test)
i = iris_cm.poof()

Desktop (please complete the following information):

• OS: macOS
• Python Version 3.6
• Yellowbrick Version 0.8

Additional context Note that the behavior appears to be somewhat different when the code is executed inside a Jupyter notebook. Discovered in the process of working on #611

[lwgray]

I am reopening this issue because during my implementation of plot directive for the ConfusionMatrix Docs #747 I came across it again. It doesn’t appear to be fixed with recent mpl updates.

• OS: Ubuntu 18.04.1 LTS
• Python Version 3.6.5
• Yellowbrick Version 0.9
• I use Anaconda
• matplotlib 3.0.2

[lwgray]

I think I might have found a partial solution (also at #768). I found that adding the keyword bbox_inches='tight' to poof fixed the cutoff.

iris_cm = ConfusionMatrix(
    model, classes=classes,
    label_encoder={0: 'setosa', 1: 'versicolor', 2: 'virginica'}
)

iris_cm.fit(X_train, y_train)
iris_cm.score(X_test, y_test)

i=iris_cm.poof(bbox_inches='tight')

Before bbox_inches='tight'

After bbox_inches='tight'