Add dead letter index for ES outputs

[emilioalvap]

What does this PR do?

Depends on #729.
Closes https://github.com/elastic/enhancements/issues/21378.

Adds es_dead_letter_index option to ES outputs, with a similar mechanism as to mimic filebeat's non_indexable_policy.

After failure to ingest on a bulk() request, ES output will now try to ingest to the provided dead_letter_index before queueing the failed event on the replay queue.

Checklist

• My code follows the style guidelines of this project
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• I have made corresponding change to the default configuration files
• I have added tests that prove my fix is effective or that my feature works
• I have added an entry in CHANGELOG.md

[gizas]

If you plan to merge it with 729 pr, please update changelog and include this PR - as per comment #729 (comment)

[constanca-m]

@emilioalvap I saw you tested the situation for when both data streams fail with a processor. Does it also happen if the dead letter index does not exist or if we pass an empty string as a value for it?

[emilioalvap]

@constanca-m As for an index that doesn't exist, that is usually handled by ES auto create policy, which tbh it's not something I've ever come across to be disabled.
Even then with auto_create disabled, my understanding is that the _bulk request will return an index not found response (404) and report it back as an error for each individual action, like a mapping error or processor rejection would. So it would behave the same way as those types of errors and end up in the replay queue.

As for the empty value, it's the default value if not defined, so it should skip the dead letter index and continue straight into replay queue, as it is now.