Memory module

[Raygama]

AURA should persevere interaction memory for at least the current session.

• Maintain in-session memory stored in Python
• Provide a function to retrieve the last N messages or interactions.
• Provides a memory to summarize for Summarizer module #6

[Raygama]

Implemented the new memory module using Supabase (pgvector) in the branch feat/stabilization-and-refinement. This replaces local Postgres/Qdrant dependencies and supports semantic search for long-term conversation history in
ai-service/app/services/memory_service.py