Implement lazy rollover for failure stores

[nielsbauman]

Allow marking failure stores for lazy rollover. When a failure occurs while trying to ingest a document, and it gets redirected to a failure store that is marked for lazy rollover, the failure store will be rolled over and the document gets redirected to the new write index of the failure store. Currently, the two places where we roll over failure stores that are marked for lazy rollover are 1. TransportBulkAction.java (failed ingest node operations) and 2. BulkOperation.java (shard-level bulk failures). Note that there is currently no automation in place for marking failure stores for lazy rollover. This is available through the rollover API, but it will most likely mainly be used by a future PR that does automate the marking for lazy rollover in some way.

[nielsbauman]

@gmarouli

I do not see why it's not a ready PR. Do you have any concerns?

I marked it as a draft as it's based off of #107562, which was still pending your review at the time of writing this :) (and I'll have to rebase this branch off of main before I can merge).

[elasticsearchmachine]

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