Make user resource usage metrics intermittent

 - Provide a lambda that will work as long as the gauge is registered
 - Produce NaN for users not using any resources (and from nodes that are not cluster leader)

Update ClusterChecker metric counters to distinguish between differen…

…t situations

Separate metrics for:
 - Host unreachable
 - Host returning a response that cannot be parsed as health status
 - Host returning status different than UP for a (non-ignored) health indicator

Publish metric counter of jobs rejected due to user exceeding their j…

…ob limit

Update agent to match server behavior when staging job dependencies

Genie V3 downloads job dependencies and configurations to the root of the job folder.
Whereas the agent was creating 'configurations' and 'dependencies' directories.

Update the agent setup code to match the server setup code.

Adjust Agent user console logging

 - Remove low-level state machine transition messages
 - Add human readable messages for major changes of state during execution

Hack: temporarily disable traversal of job dependencies sub-tree

Make job workflow beans more amenable to overrides

Job workflow auto config was missing conditionals which makes it hard to override these beans if required.

Change V3 to V4 Job Request logic to ensure no splitting of command args

For backwards compatibility with how V3 job requests behave this will lump all the v3 command args into a single element in a list without splitting them to avoid errors

Revert command arg functionality to a variation of V3 for API

This change makes the job request and job objects behave pretty much as they did in V3 with the caveat that they limit the
number of characters total in the command arguments to 10,000 and throws a precondition exception if that is exceeded.

Add health check aspect back in as a bean