Rewrite the jobs data model about rq

[nvie]

Implementation steps

Baby-steps to the final implementation:

• Get rid of the DelayedResult object (there can be a result key underneath the newly-to-create job key)
• When Jobs instantiate, generate a uuid for them. This is the job ID.
• Jobs are not actually created in Redis until they are enqueued. Before enqueuing, no Job keys will be written to Redis.
• When enqueue'ing, first write out the job into its own key, then push the job ID onto the given queue.
• When dequeue'ing, pop the job ID from the given queue. If there still exists a job key for that job in Redis, unpickle it and execute it. If not, warn about it and simply ignore the job ID.

The job key

A "job key" is a Redis hash with a key of the form rq:job:6be1faf9-69bf-4248-931a-ff6c4e37074b. The hash has the following keys:

• data: the actual payload of the job. It contains the pickled (func, args, kwargs) tuple.
• origin: the original queue the job has first been submitted to.
• enqueued_at: the last enqueue datetime (in universal time)
• created_at: the date time at which the job was first created (in universal time)
• exc_info (optional): exception information about the last failure
• result (optional): if the job returned some value, the result will be serialised (pickled) into this hash key

Future work

• To "cancel" a job, simply remove the job key.