Feature(new_metrics): migrate table-level metrics

[empiredan]

There are following metrics all of which are in class replica, replica.cpp file:

_counters_table_level_latency[]

Therefore, table-level entity can just be defined in replica.cpp, where they are also instantiated in class replica, to which the above metrics are attached.

The table-level metrics originate from https://github.com/XiaoMi/rdsn/pull/336 which is tracked by issue https://github.com/apache/incubator-pegasus/issues/406.

[acelyc111]

Now that we can aggregate on collector, is it necessary to keep and aggregate these metrics on each replica server?

[empiredan]

Now that we can aggregate on collector, is it necessary to keep and aggregate these metrics on each replica server?

OK, it's better to aggregate table metrics based on replica metrics, we can suspend this task.