Sweep all INFO level logging and aggressively drop to DEBUG, and from DEBUG to TRACE

Speaking with Arjen from Facebook ops at HBaseCon, I asked if given one single request for improving HBase operability, what would that be. The answer was to be less verbose at INFO log level. For example, with many regions opening, anomalous events can be difficult to pick out among the 5-6 INFO level messages per region deployment. Where multiple INFO level messages are printed in close succession, we should consider coalescing them. For all INFO level messages, we should be aggressive about demoting them to DEBUG level. And, since we are now increasing the verbosity at DEBUG level, the same considerations should be applied there, with coalescing and demotion of really detailed/low level logging to TRACE.

Consider making this a blocker.