Concurrency concerns in transition of space quota state with large clusters

spin-out from https://issues.apache.org/jira/browse/HBASE-16961?focusedCommentId=15985358&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15985358

As the master processes the current space usages for all tables with space quotas, it does so serially. It also synchronizes on the map which stores each table's state. This prevents the master from easily changing to process table quota violation state transition.

When we get to clusters with large number of tables, each with a space quota, we could run into a situation where a non-zero source of latency is iterating over each table in serial.