Add LSTMClassifier

[chschroeder]

Feature description

Add an LSTM Pytorch model and and LSTMClassifier.

Edit:
I am just a few weeks away from a 2.0.0-alpha release. Therefore, starting branch would be the dev branch.

Points for this task that I can see:

• Use the Pytorch LSTM implementation: https://pytorch.org/docs/stable/generated/torch.nn.LSTM.html
• small_text/integrations/pytorch/classifiers/kimcnn.py:
  • In the best case we would generalize this, to a PytorchClassifier (or TorchClassifier?).
  • The EmbeddingMixin is the most significant difference. We have to see if this part can be unified.
• The unit / integration tests would mirror the KimCNNClassifier tests

Motivation

This request has reached me via an e-mail. Also, LSTMs have been on my todo list for quite a while.

Additional comments

• Module-wise, this belongs in the pytorch integration packages.
• I won't have time for this anytime soon. Contributions are welcome.