WebHook worker logs 12k unhandled errors per week

WebHookWorker experiences ~12k unhandled errors per week:

The errors are:

EOFError

EOFError can be due to using port 443 without SSL, or transient network failures.

The SSL configuration problem is unlikely because HTTParty handles setting this, so it's like this error is due to the remote server terminating the connection.

JobReplicaNotUpToDate

Ideally this shouldn't cause an error that counts against our error budget, but for now, let's leave it as it requeues the job, which is what I assume it is intended to do.

ActiveRecord::RecordNotFound

Must happen soon after a webhook is deleted.

Errno::ENETUNREACH

Should probably be handled the same as Errno::EHOSTUNREACH is, in Gitlab::HTTP.