This marks a job as duplicate when there was already a job in the same
queue with the same arguments.

We do this by storing a key in Redis based on the argumens, worker and
queue when a job gets scheduled. If another job gets scheduled and the
key already exists, we mark the job as duplicate.

When a job starts, we delete its key from Redis.

Later we can use this for dropping jobs from redis if they are
idempotent.