cleanup: remove dead code + deduplicate GPU state reads

Dead code removed:
- globals.py: GUILD_SETTINGS (empty dict, zero consumers)
- config.py: unused 'import os'
- config_manager.py: unused 'import os' and 'Union'
- server_manager.py: duplicate 'from datetime import datetime, timedelta'

GPU deduplication:
- get_current_gpu_url() now delegates to config_manager.get_gpu()
- get_gpu_status() endpoint now delegates to config_manager.get_gpu()
- Both previously re-read memory/gpu_state.json directly

bot/api.py

@@ -72,18 +72,10 @@ api_requests_logger = get_logger('api.requests')
 # ========== GPU Selection Helper ==========
 def get_current_gpu_url():
     """Get the URL for the currently selected GPU"""
-    gpu_state_file = os.path.join(os.path.dirname(__file__), "memory", "gpu_state.json")
-    try:
-        with open(gpu_state_file, "r") as f:
-            state = json.load(f)
-            current_gpu = state.get("current_gpu", "nvidia")
-            if current_gpu == "amd":
-                return globals.LLAMA_AMD_URL
-            else:
-                return globals.LLAMA_URL
-    except:
-        # Default to NVIDIA if state file doesn't exist
-        return globals.LLAMA_URL
+    from config_manager import config_manager
+    if config_manager.get_gpu() == "amd":
+        return globals.LLAMA_AMD_URL
+    return globals.LLAMA_URL
 
 app = FastAPI()
 
@@ -697,13 +689,8 @@ def cleanup_bipolar_webhooks():
 @app.get("/gpu-status")
 def get_gpu_status():
     """Get current GPU selection"""
-    gpu_state_file = os.path.join(os.path.dirname(__file__), "memory", "gpu_state.json")
-    try:
-        with open(gpu_state_file, "r") as f:
-            state = json.load(f)
-            return {"gpu": state.get("current_gpu", "nvidia")}
-    except:
-        return {"gpu": "nvidia"}
+    from config_manager import config_manager
+    return {"gpu": config_manager.get_gpu()}
 
 @app.post("/gpu-select")
 async def select_gpu(request: Request):