fix(plugins): change log level from error to info when restarting ext…

…ernal plugin instance (#8652)

When the external plugin instances (go, python, ...) loses the instances_id in a event handler, the instance is restarted to complete the plugin call but a weird error appear:
```
2021/08/20 10:23:59 [**error**] 30#0: *51336 [kong] mp_rpc.lua:308 [jwe] no plugin instance 0, client: 10.240.0.178, server: kong, request: "POST /example HTTP/1.1", host: "example.es"
```
We change the log level to info as this is not a real problem as the plugin is restarted lately in code (considering the message string as well)

Co-authored-by: Xumin <100666470+Suika-Kong@users.noreply.github.com>

kong/runloop/plugin_servers/mp_rpc.lua

@@ -329,12 +329,12 @@ function Rpc:handle_event(plugin_name, conf, phase)
   local _, err = bridge_loop(self, instance_id, phase)
 
   if err then
-    kong.log.err(err)
-
     if string.match(err:lower(), "no plugin instance") then
+      kong.log.warn(err)
       self.reset_instance(plugin_name, conf)
       return self:handle_event(plugin_name, conf, phase)
     end
+    kong.log.err(err)
   end
 end
 

kong/runloop/plugin_servers/pb_rpc.lua

@@ -394,13 +394,13 @@ function Rpc:handle_event(plugin_name, conf, phase)
     event_name = phase,
   }, true)
   if not res or res == "" then
-    kong.log.err(err)
-    if    string.match(err:lower(), "no plugin instance")
-       or string.match(err:lower(), "closed")  then
-
+    if string.match(err:lower(), "no plugin instance")
+      or string.match(err:lower(), "closed")  then
+      kong.log.warn(err)
       self.reset_instance(plugin_name, conf)
       return self:handle_event(plugin_name, conf, phase)
     end
+    kong.log.err(err)
   end
 end