request logger retries

[ryandawsonuk]

The request logger currently supports retries on conflicting inserts/updates - https://github.com/SeldonIO/seldon-core/blob/master/components/seldon-request-logger/app/default_logger.py#L133

But not for the case of a connection blip. These should be rare but I've seen them happen occasionally during a load test on openshift (using the openshift cluster logging operator's install of elastic). Did use python retry library in a previous incarnation. I don't think we want retries by default as that could lead to request numbers multiplying but it would be good to have the option to set (which we could then enable for this particular openshift setup).

[ryandawsonuk]

Actually just tried this again and it was fine. I may have been getting confused before by Unverified HTTPS request warnings. Don't think we need this after all.