Calls to oracle should respond to 502/504 errors with longer timeouts

[seasidesparrow]

Currently, docmatch pipeline will respond to a 50x error returned by oracle service by waiting ten seconds and trying again, up to a limit ten retries. The service and/or the Solr searchers can get into states where they generate multiple 50x errors, leading to multiple attempts in a (relatively) short period of time. These states can last from minutes to hours, so giving an oracle call what amounts to a 100-second window to resolve is probably too short.

We should increase the timeouts as the number of timeouts increases, e.g. 10 seconds, 20 seconds, 30 seconds... or 10 seconds, 100 seconds, 1000 seconds. A linear scaling (sum(n*10)n=1,10) would lead to a 550-second window