perf: Downgrade model to 1.5B for faster CPU inference

perf: Optimize Latency (Remove query translation, async cache, tune LLM threads, limit tokens)

feat: Phase 4 Upgrade - RAG Architecture Stability (mMARCO, pdfplumber, true recursive chunk, 3B LLM)

feat: Implement Phase 3 - Optimize inference engine with llama.cpp

feat: production-hardening, upload validation, admin auth, threadpool, logging & HF dataset repo integration

Expand LLM context window max length to 6500 tokens (~5000 words)

Use native Transformers GGUF loader to fix C++ compile delay in Docker build

Upgrade Local LLM to GGUF (llama.cpp) format for 5x-7x faster CPU inference

Add accelerate dependency and fix Local LLM loading error

DocBrain RAG official release for HF Space