Add support for "can execute" logic

[adamkewley]

Some jobs might require clusters/services, which go down from time to time. Instead of running the job, some kind of provision for declaring that something can be executed would be useful - although this might require background polling.

[adamkewley]

Although, thinking on it (for one minute), this might be something implementers should do in their downstream scripts (e.g. while (cluster is down) { sleep(some_period); }). This would be more transparent than having a whole extra pipeline + poller, etc.