[dGPU] Use cldnn concat if data type is fp32. (#19240)

src/plugins/intel_gpu/src/graph/layout_optimizer.cpp

@@ -1594,7 +1594,8 @@ impl_types layout_optimizer::get_preferred_impl_type(program_node& node, format
         if (!_optimization_attributes.use_onednn_impls)
             return impl_types::ocl;
 
-        if (node.get_output_layout().data_type == data_types::i32)
+        if (node.get_output_layout().data_type == data_types::i32 ||
+            node.get_output_layout().data_type == data_types::f32)
             return impl_types::ocl;
 
         for (auto& dep : node.get_dependencies()) {