Checking failure store used for wrong data_stream.type from 9.2.0

[eyalkoren]

A followup of #131261 - starting in 9.2.0, the failure store will be enabled by default for logs-*-* data streams, so sending the wrong data_stream.type doesn't cause 400s anymore, but rather 201 responses with failure_store: used. This causes failures in the compatibility tests, hence this PR.

I couldn't find a way to add another test - Test wrong data_stream type - logs before 9.2.0 (e.g. to that 8.19 builds keep verifying the current behavior), because I don't know how to restrict a REST yaml test from running up to a specific version.

@elastic/es-core-infra please review the technical aspect of this.

I am still looking for approval for the conceptual issue of whether stop responding with errors can be considered a non-breaking change by default.

[elasticsearchmachine]

Pinging @elastic/es-data-management (Team:Data Management)

[dakrone]

I don't think this is normally how we'd do this. Instead, we would add either a "capability" to the endpoint, or else limit where the test runs when run from 9.2.

[eyalkoren]

I don't think this is normally how we'd do this. Instead, we would add either a "capability" to the endpoint, or else limit where the test runs when run from 9.2.

What do you mean by "adding a capability to the endpoint"? Do you mean that the index/bulk APIs check whether the client is of a specific version AND whether the failure store is enabled and baed on that decide what to respond?
As for limiting the test - I think the test of not accepting wrong data_stream.type is valid and useful, only it needs to be different when the failure store is on. WDYT?