Fix Trash migration

Since NXP-24031, the trash state is now a dedicated boolean instead of the lifecycle state deleted.

We implemented a migrator to use new service along with NXP-24035. This migrator retrieves all document whose lifecycle state is deleted and sets the trashed boolean to true.
The migrator initially doesn't change the lifecycle state for several reasons:

• transition to leave deleted state is not necessarily known
• lifecycle policy can no longer exist

Since NXP-25439, server is able to auto-detect in which case the DB is. This allows to easily switch from a former implementation to a knew one without contributing and/or restarting the server.

As deleted state is left as it on document, server will detect DB is in former state, which can bring several issues. Especially if nodes in a HA setup don't use the same service.

We need to fix this:

• a simple solution could be to follow the undelete transition
• if not possible, hardly set project state
• also change the backward compatibilty mechanism on CoreSession#followTransition, as having documents with deleted state cause migration service to switch trash service to former one! -> change backward compatibilty mechanism to just trash document and not follow transition in such case

Note: NXQL queries are rewritten at runtime depending on trash service state. We can choose to leave the query as is when constraints contain both lifecycle state and trashed property.