feat: Variet Engine v1.0 + 5-model tuning complete

Phase 01 (LLM Tuning):
- Gemma4 26B: 74.65 t/s (fast)
- Qwen 35B: 61.62 t/s (balanced)
- Gemma4 31B: 16.0 t/s (deep-coder)
- Qwen 27B: 16.7 t/s (deep-logic)
- Qwen 122B: 8.95 t/s (ultra, GPU 1 only)

Phase 02 (API Engine):
- FastAPI reverse proxy on port 8000
- /engine/switch hot-swap with 503 protection
- config/engine_models.json as single source of truth
- Replaced 4 individual .bat files with unified engine

File cleanup:
- scripts/ 85 files -> 9 + _archive/
- Root .bat files -> _archive/

_archive/start_qwen_122b_optimized.bat

@@ -0,0 +1,29 @@
+@echo off
+chcp 65001 >nul
+echo =========================================================
+echo  Qwen3.5 122B-A10B API Server (Final Optimized)
+echo  [CORE] --cpu-moe + -ngl 999: WDDM swap 완전 회피
+echo  [TUNED] -t 4: CPU 캐시 경합 최소화 (10.83 t/s)
+echo  [TUNED] mmap on: 메모리 매핑 활성 (no-mmap보다 빠름)
+echo  [VRAM] 5.4GB / 12GB (54%% 여유)
+echo =========================================================
+echo.
+
+llama_bin_run\llama-server.exe ^
+  --model models\Q4_K_M\Qwen3.5-122B-A10B-Q4_K_M-00001-of-00003.gguf ^
+  -ngl 999 ^
+  --cpu-moe ^
+  -c 2048 ^
+  -np 1 ^
+  -fa on ^
+  --cache-type-k q4_0 ^
+  --cache-type-v q4_0 ^
+  -ub 256 ^
+  -b 1024 ^
+  -t 4 ^
+  --mlock ^
+  --prio 2 ^
+  --port 8000 ^
+  --host 0.0.0.0
+
+pause