Fix multi-threading issues (#2013)

- Implement atomic.fence to ensure a proper memory synchronization order
- Destroy exec_env_singleton first in wasm/aot deinstantiation
- Change terminate other threads to wait for other threads in
  wasm_exec_env_destroy
- Fix detach thread in thread_manager_start_routine
- Fix duplicated lock cluster->lock in wasm_cluster_cancel_thread
- Add lib-pthread and lib-wasi-threads compilation to Windows CI

core/iwasm/interpreter/wasm_runtime.c

@@ -2073,6 +2073,15 @@ wasm_deinstantiate(WASMModuleInstance *module_inst, bool is_sub_inst)
     if (!module_inst)
         return;
 
+    if (module_inst->exec_env_singleton) {
+        /* wasm_exec_env_destroy will call
+           wasm_cluster_wait_for_all_except_self to wait for other
+           threads, so as to destroy their exec_envs and module
+           instances first, and avoid accessing the shared resources
+           of current module instance after it is deinstantiated. */
+        wasm_exec_env_destroy(module_inst->exec_env_singleton);
+    }
+
 #if WASM_ENABLE_DEBUG_INTERP != 0                         \
     || (WASM_ENABLE_FAST_JIT != 0 && WASM_ENABLE_JIT != 0 \
         && WASM_ENABLE_LAZY_JIT != 0)
@@ -2155,9 +2164,6 @@ wasm_deinstantiate(WASMModuleInstance *module_inst, bool is_sub_inst)
     wasm_externref_cleanup((WASMModuleInstanceCommon *)module_inst);
 #endif
 
-    if (module_inst->exec_env_singleton)
-        wasm_exec_env_destroy(module_inst->exec_env_singleton);
-
 #if WASM_ENABLE_DUMP_CALL_STACK != 0
     if (module_inst->frames) {
         bh_vector_destroy(module_inst->frames);
@@ -2241,7 +2247,6 @@ call_wasm_with_hw_bound_check(WASMModuleInstance *module_inst,
     WASMRuntimeFrame *prev_frame = wasm_exec_env_get_cur_frame(exec_env);
     uint8 *prev_top = exec_env->wasm_stack.s.top;
 #ifdef BH_PLATFORM_WINDOWS
-    const char *exce;
     int result;
     bool has_exception;
     char exception[EXCEPTION_BUF_LEN];