Suggesting fixes for syntax errors using sensibility

[alexmojaki]

Continuing from https://github.com/alexmojaki/futurecoder/issues/53#issuecomment-698454058

In a nutshell for newcomers: https://github.com/naturalness/sensibility suggests fixes for syntax errors with machine learning, but training a Python model requires a lot of computation power. A pre-trained model for Python is not currently available.

1. Do you have pre-trained models for any languages? It'd be nice just to be able to play with the code directly.
2. Did you actually train and evaluate the model for Python at some point, or was that just some unfinished work in the repo?
3. Can you give an idea of what the computational cost of the LSTM is like for prediction once the model has already been trained?

This idea has become less relevant now that we have friendly-tracebacks suggesting fixes for many common syntax errors.