Update the RAG Knowledge Assistant entry for semantic and hybrid retrieval #2

Context

rag-knowledge-assistant now offers TF-IDF, semantic (Qdrant + fastembed) and hybrid (Reciprocal Rank Fusion) retrieval, selectable by configuration, and measures each mode on a 40-question golden set with a CI quality gate (LacerdaTraderCode/rag-knowledge-assistant#4). Two statements in this repository no longer describe the project.

Scope

• README.md, RAG Knowledge Assistant section: update the stack line, and replace the bullet saying TF-IDF was chosen over embeddings.
• ROADMAP.md, Applied AI table: update the stack and what the RAG row demonstrates.
• Use only numbers the project's evaluation produced. Keep the existing tone and formatting.

Done when

• Neither file says TF-IDF is the only retrieval method or that it was chosen instead of embeddings.
• The figures quoted match the README of rag-knowledge-assistant.