#4 Separate <run> and <start> functions

[Austin Hastings]

Presently <run> is documented as starting a job, and then returning a status based on the success/failure of the act of running the child job.

The only way to "call" a job is to use the <depends> task.

I suggest separating <run> and <start> functions, with <start> behaving as <run> presently behaves, while <run> (or perhaps <call>) returns the result of its child, or an exception if the kickoff fails.