Implement BiRefNet and HAT Sharper models with background removal capabilities

- Added validation documentation for AI model tiers.
- Introduced `background.py` for background removal logic, including inference functions for alpha and foreground extraction.
- Created `onnx_birefnet.py` to handle BiRefNet model sessions and alpha inference.
- Developed tests for background models, ensuring proper dispatch and session reuse.
- Added tests for BiRefNet ONNX model to validate alpha output and error handling for invalid values.
- Implemented model export scripts for BiRefNet and HAT Sharper, ensuring proper weight compression and licensing.
- Created `deform_sampling.py` for deformable convolution operations used in BiRefNet.
- Added utility functions for compressing ONNX weights without altering graph arithmetic.
- Introduced a reference plane node for depth field adjustments in geometry processing.

tests/anyimage/server/test_model_inference.py

@@ -93,7 +93,7 @@ class ModelInferenceTest(unittest.TestCase):
 
 upscale = load_server_module("models.upscale")
 
-@pytest.mark.parametrize("key", ["REALESRGAN_GENERAL_WDN_X4V3", "SAFMN_REAL_X4", "REALESRGAN_X4PLUS"])
+@pytest.mark.parametrize("key", ["REALESRGAN_GENERAL_WDN_X4V3", "HAT_GAN_X4_SHARPER", "REALESRGAN_X4PLUS"])
 def test_upscale_models_resize_prediction_then_restore_alpha(tmp_path, key):
     rng = np.random.default_rng(1)
     source = Image.fromarray(rng.integers(0,256,(5,7,4),dtype=np.uint8))
@@ -101,7 +101,8 @@ def test_upscale_models_resize_prediction_then_restore_alpha(tmp_path, key):
     source.save(path)
     prediction = Image.fromarray(rng.integers(0,256,(20,28,3),dtype=np.uint8))
     manager = SimpleNamespace(directory=lambda _: tmp_path, get_upscale=lambda *args: object())
-    with patch("server.models.onnx_upscale.infer", return_value=prediction):
+    with patch("server.models.onnx_upscale.infer", return_value=prediction) as infer:
         output = upscale.infer_one(manager, {"model":key,"device":"cpu"}, path)
+    assert infer.call_args.kwargs["tile_border"] == (48 if key == "HAT_GAN_X4_SHARPER" else 16)
     np.testing.assert_array_equal(output.convert("RGB"), prediction.resize((14,10), Image.Resampling.LANCZOS))
     np.testing.assert_array_equal(output.getchannel("A"), source.getchannel("A").resize((14,10), Image.Resampling.LANCZOS))