separate load functionality from inference client to loader client. Call loader client from inference to get the model.

remove dummy dlls, remove resource loader from c#.

TODO: Load dlls separately by Loader UI and loader client

WIP

Azaion.Loader/hardware_service.pyx

@@ -1,34 +1,9 @@
 import os
 import subprocess
-import pynvml
-
-
 cdef class HardwareService:
 
     @staticmethod
-    cdef has_nvidia_gpu():
-        try:
-            pynvml.nvmlInit()
-            device_count = pynvml.nvmlDeviceGetCount()
-
-            if device_count > 0:
-                print(f"Found NVIDIA GPU(s).")
-                return True
-            else:
-                print("No NVIDIA GPUs found by NVML.")
-                return False
-
-        except pynvml.NVMLError as error:
-            print(f"Failed to find NVIDIA GPU")
-            return False
-        finally:
-            try:
-                pynvml.nvmlShutdown()
-            except:
-                print('Failed to shutdown pynvml cause probably no NVidia GPU')
-                pass
-
-    cdef str get_hardware_info(self):
+    cdef str get_hardware_info():
         if os.name == 'nt': # windows
             os_command = (
                 "powershell -Command \""