generic alert when application process a block (or batch of blocks) to slow

There are situations when during live sync an application processes single blocks too slow (longer than 3s), and in consequence, loses sync with the chain head block. We could add to the HAF generic mechanism which will measure how much time takes one iteration of an application, and if it is longer than 3s then raise a warning and add some information to some statistic table (block and time).

Because some applications may process blocks longer than 3s (i.e. run a long task from time to time for a specific custom JSON operation) the alarm should be optional (i.e. time observation for a given context can be started/stopped by the app).

Such an alarm can also be applied to stages other than 'live'.(option)