Add web search demo temporarily as a py file

[pratyaypandey]

As we navigate issues setting up a Colab for this demo, I still wanted to get feedback on the text (represented through the multiline comments) and substance of the demo. Later, I will port this into the Colab.

The deep research demo attempts to be iterative, with each example building onto the previous one. First, we simply use web_search and sem_agg with Google. Later, we integrate the arXiv search, too.

This provides the initial two streams of information. To demonstrate the scale of this search functionality, we then grab 50 results from them, and use sem_topk to filter to the most relevant for our answer.

Finally, we make use of parse_pdf to also integrate textbook data towards our answer.

Hopefully, in going through the demo, a reader gains intuition into LOTUS, understanding how to iteratively scale projects with LOTUS.