[Agentic AI][LLM] Standardize model artifact preflight for benchmarks

[makr-code]

Problem: LLM/RAG/LoRA benchmarks fail due to missing artifacts (for example gguf model files).\n\nGoal: implement benchmark preflight checks and setup workflow.\n\nAcceptance:\n- Preflight validates required model artifacts before run\n- Missing artifact errors are converted to actionable setup guidance