nilnon

Now the elasticsearch index name suffix is based on `execution_date`. In the scenario of task rerun, many historical indexes will be generated. Can this rule be optimized?

The `task_instance` and `task_datum` tables can be optimized into partitioned tables.