[GPU] Reduce limitation for choosing onednn deconvolution (#24960)

### Details:
- Remove feature depth limitation in is_node_for_onednn() for
deconvolution

### Tickets:
 - 136045

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

@@ -905,13 +905,12 @@ static bool is_node_for_onednn(deconvolution_node const& node) {
     bool onednn_valid_dt = layout_optimizer::are_data_types_suitable_for_onednn((program_node&)node);
 
     bool onednn_valid_params = onednn_valid_dt &&
-                               input_layout.feature() >= 16 &&
                                prim->groups == 1 &&
                                get_post_ops_count(node) <= 32;
 
     auto spatial_dims_num = input_layout.get_spatial_rank();
 
-    return onednn_valid_dt && onednn_valid_params && spatial_dims_num <= 3;
+    return onednn_valid_params && spatial_dims_num <= 3;
 }