Example for LLM model

[a-r-r-o-w]

As mentioned in #34, this example contains:

• An example to highlight how UpTrain can be used to smartly fine-tune a LLM
• A model signal using sentiment analysis
• A model signal using keyword search

This PR is opened as a draft because of the following TODOs (please add on others that are required):

• Improve code documentation
• Add uptrain logging to create plots showing loss and perplexity on Streamlit dashboard
• Add visualizations for things like weight of prediction of different words to provide a graphical representation of what the model thinks is the best masked word

Edit: I've added loss, perplexity and mask prediction score visualizations using the streamlit dashboard functionality.