Web search performance improvements

[zavocc]

Requires chroma server running! Configurable through dev.env

• Webpages fetched from duckduckgo results now chunks them and finds the most relevant information based on chunks

• Now uses chroma along with batch embedding them asynchronously for finding key points

• Performance with models should be bit faster if not quite

• [x] Utilize embeddings as part of workflow

• [x] Use chroma server asynchronously

• [x] Adjust chunk size and use basic chunker

• [x] Test the perfoamnce and parallel execution by different users

• [x] Update docs (dependencies and web search internals)