fix(core): Reduce logging overhead for levels that do not output (#7479)

all current logging calls execute `callsites()` to figure out what code
tried to log. This happens even for logging methods that aren't supposed
to create any output. Under moderate load, this can take up quite a lot
of resources. This PR changes the logger to make all ignorable logging
methods a No-Op.

In a small benchmark with a simple webhook, with log-level set to
`warn`, and using `ab -c 50 -n 500
http://localhost:5678/webhook/testing`, these were the response times:

### Before

![Before](https://github.com/n8n-io/n8n/assets/196144/01680fd9-3d2a-4f7f-bb1c-5b03bd7d5bc3)

### After

![After](https://github.com/n8n-io/n8n/assets/196144/ccacb20a-48ca-455a-a8cb-098c9c0e352e)

packages/workflow/src/LoggerProxy.ts

@@ -22,21 +22,21 @@ export function log(type: LogTypes, message: string, meta: object = {}) {
 // Convenience methods below
 
 export function debug(message: string, meta: object = {}) {
-	getInstance().log('debug', message, meta);
+	getInstance().debug(message, meta);
 }
 
 export function info(message: string, meta: object = {}) {
-	getInstance().log('info', message, meta);
+	getInstance().info(message, meta);
 }
 
 export function error(message: string, meta: object = {}) {
-	getInstance().log('error', message, meta);
+	getInstance().error(message, meta);
 }
 
 export function verbose(message: string, meta: object = {}) {
-	getInstance().log('verbose', message, meta);
+	getInstance().verbose(message, meta);
 }
 
 export function warn(message: string, meta: object = {}) {
-	getInstance().log('warn', message, meta);
+	getInstance().warn(message, meta);
 }