Status Json Generation Can Be Very Long

`clusterGetStatus()` does many steps in serial order. Even though each step typically has a timeout, the total time can be very long, especially when there are faults in the cluster, e.g., when some storage servers are unavailable.

A second problem is that if a new step is added without a proper timeout, the process becomes unbounded. This is bad, because operational tools typically depends on the output from `status json`.

To bound the time for status generation, we need to parallelize the steps for generating status as much as possible, probably with a good code refactoring to address both of the above problems.