[GPU] enable dynamic shape conv on compile graph pass (#30944)

### Descriptions
- Remove WA that checks use_explicit_padding for convolution. It needs
to choose impl at compile graph pass.

### Tickets:
 - 162880

Author: sungeunk

src/plugins/intel_gpu/src/graph/registry/convolution_impls.cpp

@@ -23,9 +23,7 @@ const std::vector<std::shared_ptr<cldnn::ImplementationManager>>& Registry<convo
         OV_GPU_CREATE_INSTANCE_OCL(ocl::ConvolutionImplementationManager, shape_types::static_shape)
         OV_GPU_CREATE_INSTANCE_OCL(ocl::ConvolutionImplementationManager, shape_types::dynamic_shape,
             [](const cldnn::program_node& node){
-                if (node.can_use(impl_types::onednn))
-                    return false;
-                return node.as<convolution>().use_explicit_padding();
+                return !node.can_use(impl_types::onednn);
         })
     };