Search engine schema update documentation // migration

[fschade]

updating the search engine schema right now is a painful process which could include re-indexing the whole instance, making it batteries included is not an option in my opinion because we have no insights what each individual installation contains and what were allowed to touch or not.

Because of this we have to document one or two cases how such an schema update could look like, what needs to be done and evaluate feasible ways of a non re-indexing workflow.

This needs to be done for:

• bleve search (our default build in search engine)
• openSearch

[dschmidt]

We have an explicit message, that opensearch index is incompatible.
Can/should we have an explicit message for bleve as well at least?
I would like to avoid silent failure or an unreadable (and "unactionable" error message).

I agree that a batteries included is very hard to get right, especially when thinking of kubernetes environments with multiple instances of the backend: which instance is allowed to update?