Add query functionality to the Document Loader notebook

[averikitsch]

Currently, the Document Loader notebook only outlines how to use the "table_name" argument to pull all documents. We need to highlight that the Loader can be used with "query" to load documents. This should also note how the columns get arranged into the document's content vs metadata.

We should also add a note to the vector store notebook that this could be a way to explore filters for similarity search or just search documents directly without embeddings.