RnnSequenceClassifier has null model after deserialization

[ooogood]

Describe the bug
RnnSequenceClassifier's member variable "isInitializationFinished" is not set at the end of member method "initializeClassifier(Instances data)".
Similar to the Dl4jMlpClassifier::initializeClassifier(Instances data), it should call Dl4jMlpClassifier::finishClassifierInitialization() at the end.

Otherwise, the "isInitializationFinished" flag will remain false, and the deserialized RnnSequenceClassifier object will have a null model. (Because in Dl4jMlpClassifier::writeObject, if the flag is not true, the model won't be serialized into the object file.)

To Reproduce
Steps to reproduce the behavior:

1. Train a RnnSequenceClassifier like in this page: https://deeplearning.cms.waikato.ac.nz/examples/train_rnn/
2. Serialize the classifier with SerializeHelper.write( "path/to/object/file", classifier)
3. Deserialize the classifier with (RnnSequenceClassifier)SerializeHelper.read( "path/to/object/file" )
4. The model in the classifier should be null.

Expected behavior
The model in the classifier should be the same as before the serialization.

Screenshots

Additional Information

• Weka version: 3.8.6
• wekaDeeplearning4j package version: v1.7.2
• Operating System: Windows

Additional context