Add load forecasting example using LSTM. Add LSTM implementation consistent with blocks interface

[alpargun]

• This PR adds a load forecasting example using an LSTM model trained on a public dataset. Neuromancer's Problem formulation and Trainer are used to train the network, and the performance metrics and forecast plots are included for the evaluation. Both a Python and a Jupyter notebook file are included.
• Additionally, adds an LSTM implementation that complies with Neuromancer's blocks interface in blocks.py

[RBirmiwal]

Hi @alpargun for your contribution to NeuroMANCER! This is a great notebook. We will take a closer look how it fits into our release schedule and backlog and hopefully will incorporate into library soon. Thank you

[drgona]

@alpargun, nice job on the example. I have a few questions and suggestions.

• why do you need to create a custom dataloader? As far as I can see you could use our default DictDataset class.
• The notebook would be more didactic if it contained more markup, motivated the example, described the dataset, visualized the data, explained the model choice, and included a discussion of results. If some of these descriptions are already made in previous notebooks in the library you can reference them in the notebook.

[alpargun]

@drgona Thank you for the feedback.
I will update my example using the DictDataset class, and with more markup to explain the points in your suggestions.