Add job tracking across the cluster

[nevali]

Define APIs which can be used by libcluster applications for creating and updating information in the cluster database (or equivalent) reflecting their processing state.

Jobs have:

• An ID, which may be application-defined, but will be in the form of a UUID
• A cluster key and environment
• An optional parent
• A status
• A current instance ID (or NULL if a job isn't currently being processed)
• Creation and update timestamps
• Optional key-value metadata pairs for storing information useful for monitoring applications (such as source and destination URIs, for example)
• Event log messages (which may or may not be stored in the cluster database)

Child jobs can be used where multiple instances should process parts of a job in parallel.