fix(jit/cpu): make lar -enable_jit args possible

on other backends, for example cpu with MLIR

GitOrigin-RevId: ef98d46be326417c85fcdc482f02758e352e2f52

lite/src/mge/network_impl.cpp

@@ -79,11 +79,6 @@ void NetworkImplDft::application_config() {
             force_output_use_user_specified_memory,
             force_output_use_user_specified_memory);
     ConfigOption(no_profiling_on_shape_change, no_profiling_on_shape_change);
-    LITE_ASSERT(
-            m_user_config->options.jit_level == 0 ||
-                    (m_user_config->options.jit_level > 0 &&
-                     device_type == LiteDeviceType::LITE_CUDA),
-            "jit only support in cuda device.");
     ConfigOption(graph_opt.jit, jit_level);
     ConfigOption(comp_node_seq_record_level, comp_node_seq_record_level);
     ConfigOption(graph_opt_level, graph_opt_level);

src/core/include/megbrain/graph/cg.h

@@ -366,7 +366,6 @@ public:
          *    construction
          * 2: level-2: level-1, plus global optimization before graph
          *    compiling
-         * 3: also enable JIT
          * <0: corresponding level, with result check for debug
          */
         int16_t graph_opt_level = 2;