Allow persistent/cluster-wide setting of common data service parameters

More and more we have the need to leverage the tunables provided by our data service engine(s).  Currently, the most common ones are done through 'cbepctl' and are per-node and non-persistent.

While useful for on-the-fly diagnostics, it creates a challenge when these settings are needed for the proper functioning of certain use cases:

• They need to applied via a non-standard CLI process which makes them hard to set externally
• A restart of the node or process reverts them back to the original value and so need to be re-applied even in the event of a crash
• New nodes coming into the cluster do not inherit these settings and so again requires specific intervention when scaling out as well as in some upgrade scenarios.

This is a request to move these commonly used parameters over to ns_server to address the above concerns.  I would suggest that no change is actually needed to the way 'cbepctl' behaves.

The common settings are:

• access_scanner_enabled
• exp_pager_stime
• max_num_readers
• max_num_writers
• max_num_shards
• max_num_nonio
• warmup_min_memory_threshold
• mem_low_wat
• mem_high_wat