Add ONNX model inference implementations for MoGe-2, BEN2, and DA3

- Introduced `onnx_moge2.py` for MoGe-2 ONNX inference with NumPy post-processing.
- Added `onnx_ben2.py` for BEN2 ONNX inference with Pillow and NumPy preprocessing.
- Created `onnx_da3.py` for Depth Anything 3 inference with ONNX Runtime and NumPy.
- Implemented shared ONNX Runtime session helpers in `onnx_runtime.py`.
- Developed unit tests for server functionalities in `test_server.py`, including model caching and job management.
- Enhanced job server to handle frame extraction without requiring models.

tests/test_da3_onnx.py

@@ -12,11 +12,11 @@ SOURCE_DIR = (
     Path(__file__).resolve().parents[1]
     / "src"
     / "snapform"
-    / "ai_service"
+    / "server"
 )
 sys.path.insert(0, str(SOURCE_DIR))
 
-import da3_onnx
+import onnx_da3
 
 
 class FakeSession:
@@ -37,7 +37,7 @@ class FakeSession:
 class Da3OnnxTest(unittest.TestCase):
     def test_target_size_uses_patch_multiple(self):
         self.assertEqual(
-            da3_onnx.target_size((800, 600), 504, "upper_bound_resize"),
+            onnx_da3.target_size((800, 600), 504, "upper_bound_resize"),
             (504, 378),
         )
 
@@ -47,7 +47,7 @@ class Da3OnnxTest(unittest.TestCase):
             Image.new("RGB", (80, 60), (10, 20, 30)).save(image_path)
             session = FakeSession()
 
-            result = da3_onnx.infer(
+            result = onnx_da3.infer(
                 session,
                 [image_path],
                 process_res=56,