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  ### After 