EventIDRage c'tor: Don't reset UNDEF values to 0

Following the discussions on https://its.cern.ch/jira/browse/ATLASRECTS-7086 and https://its.cern.ch/jira/browse/ATR-25671.

Up to now the constructor of EventIDRange set the start-timestamp to 0 for Run/Lumiblock-indexed conditions and the start-Run/LB to 0 for timestamp-indexed conditions. This creates problems with they way ATLAS deals with mixed-range conditions (eg conditions depending on both the Run/Lumiblock and time-stamp).

This MR removes these lines of code, so the values in the EventIDBase object defining the start of the range remain UNDEFNUM.