Improve resilience of Nuxeo in case of infrastructure failures, namely ES erratic response times

When the audit is stored in ElasticSearch, if ES exhibits out of a sudden very large response times (1000+ times larger than usual), this can break the Nuxeo internal audit pipeline. Restoring the audit pipeline requires a Nuxeo restart.

The request is to fix this behavior so that the pipeline can restore on its own without requiring a full Nuxeo node restart.

The fact that Nuxeo audit can stop working has consequences e.g. with Drive clients synchronization mechanism, which can be stopped as a consequence, making this a very visible failure.