Gatling simulation should not require long ELB timeout

Today Gatling simulations are using synchronous endpoint to measure processing time.

This is the case for:

• the importer where we wait for the end of the sync part
• after write operation where we wait for async processing like audit, works and indexing using Elasticsearch.WaitForIndexing automation operation
• bulk command using Bulk.WaitForAction automation operation
• Re-indexing

These automation operations are only for testing and they are an easy way to measure time from the client-side without having to poll regularly the server.

On the other hand, it requires that ALB and apache timeout be configured to support a long timeout (1h) which is not always possible.

We should estimate how hard it is to have a simulation that requires no change in the target environment.