Remove data flow dumps or make them configurable
Redundant copies of the dumpDataFlow() output are produced when the overall output level for a job is set to DEBUG, see https://its.cern.ch/jira/browse/ATR-21450. This makes the log files harder to search through, and makes the data dependency information harder to understand (since a user might incorrectly assume that a dependency had been duplicated).
Removed the PrecedenceRulesGraph::initialize() dump, and made the PrecedenceSvc::initialize() dump conditional on a configuration parameter that defaults to false.