Ability to fork prior to performing a given job to increase mainline worker stability

[iansltx]

There are a few job types that we've been having intermittent issues with, primarily due to timeout errors on third party services, which will hang/terminate the worker despite having exception handling. If we could set a flag within the job object so that those jobs would fork before executing, their failure would be less destructive to the overall worker process.

Granted, this is a bit of a major change to the library, but it should be doable without too much more code. If we don't convert to another queueing system soon I'll implement this on my fork.

[josegonzalez]

Feel free to implement this change and I'll merge it in.