Purge rogue and stale entries from history set

A database exception can while confirming an END_MISS event in 2 scenarios:
1) sla end_miss for workflow_action occurring way before action was even inserted into Oozie DB, because of the sla timings provided by user
2) rogue job in memory map, but actually missing from database

Such an exception is currently handled to continue with the end_miss event, to proceed with the processing. Removing an entry from tracking, on exception, only caters to case 2 but not case 1 (valid). So something like a purge should be done on SLA Calculator Memory history set to get rid of any rogue entries, and free up resources.