PyTorch implementation of MQ-CNN and MQ-RNN

[lostella]

References:

• https://arxiv.org/pdf/1711.11053.pdf
• MXNet implementation (in my opinion this is a little too dispersive)

Milestones for this one:

• Implementing core modules (with option for different encoders, but more compact than the MXNet one)
• Implementing LightningModule to handle training: this should take care of the "forking" logic described in the paper
• Estimator wrapper

[baharian]

Any updates on this?