Improve indexing reliability when Elastic is overloaded

During document indexing with the WorkManager when Elastic is overloaded,
we have different types of errors raised by the elastic:

• indexing queue rejection
• circuit breaker
• 429 returned by AWS elastic

As the result, the indexing work is either completed with failures or put in the DLQ and this creates a discrepancy between what is indexed and the repository content.

Instead, a retry with a long delay should be performed to reduce the pressure.

Another option could be to schedule a reindexing bulk command but this fallback processing may be too complex and it is going to be hard to evaluate if it works.