Easy networks validation splitting keyword

[cnellington]

Reproducable with for v0.2.4

import numpy as np
from contextualized.easy import ContextualizedCorrelationNetworks
C_train = np.random.normal(0, 1, (100, 20))
X_train = np.random.normal(0, 1, (100, 20))
model = ContextualizedCorrelationNetworks(**model_kwargs)
model.fit(C_train, X_train, val_split=.5)

Pre-training kwarg validation gives "Received unknown keyword argument val_split, probably ignoring."

[blengerich]

Fix in #161