OCT-169: fix migration of ES index

[tseho]

Context

The migration is executed asynchronously.
Before the doctrine migration was executed, the webhook consumer started indexing new documents in the index akeneo_connectivity_connection_events_api_debug, with a new property id.
Because you can index documents without configuring a mapping beforehand, the new property has been automatically created by ES with the type text instead of keyword.

In ES, you cannot directly change the type of a property. (see https://www.elastic.co/guide/en/elasticsearch/reference/current/indices-put-mapping.html#updating-field-mappings)

The migration is failing for customers having sent webhooks between the deployment and the doctrine migration execution.

The dirty solution

Since we cannot rename a property, cannot change the type of a property, cannot remove a property of an existing index without re-indexing everything, and we cannot re-index this particular index.

Here what's happening in the new migration:

• if the id property already exists
• then, creates a temporary indice
• copy documents from the corrupted indice to the temporary one (using re-index with source + dest)
• redirects the alias from the corrupted indice to the temporary one (so what we are still writing/reading somewhere)
• delete the corrupted indice
• create a new indice with the correct mapping and the original name
• copy documents from the temporary indice to the new one
• redirects the alias from the temporary indice to the new one
• removes the temporary indice

Definition Of Done (for Core Developer only)

• Tests
• Migration & Installer
• PM Validation (Story)
• Changelog (maintenance bug fixes)
• Tech Doc