Refactor app to have a LLM models selection

Update app

Use Cohere's Rerank to improve search retrieval performance

Update document_retriever.py

Refactor app

Improve performance with contextual compression, a technique where retrieved documents are compressed, and irrelevant information is filtered out.

Update document_retriever.py

Disable tokenizer transformer parallelism to avoid deadlocks

Add new app icon. Refactor and cleanup.