Feat/restart parent

[loicmathieu]

This PR allows to restart a parent execution when we change the status of a subflow execution task that fail (which restart the subflow execution).

It was not the case before because a parent execution is notified of subflow execution thanks to the subflow execution storage (a table in JDBC and a store in Kafka) which is deleted when the subflow execution ends. So when a subflow execution restart, there are no more subflow execution inside the storage and the parent is never notified again.

Th implementation was changed so the subflow execution storage is no longer needed, we use the information of the execution trigger to know that we need to notify a parent flow so it can be restarted.

The underlying table is not deleted to avoid potential issues during migration; a follow-up issue will be created so the table will be removed in 0.22.

The PR also introduced some small changes to improve change state and restart:

• Change state was sometimes referred as changed status, this is now aligned everywhere
• Restarting due to change state now adds the system.restarted: true system label

This PR introduced a (or hit an existing) race in ForEachItem that must be fixed before we can merge it