Add potential fix to prd association bug ATLIDTRKCP-274

ATLIDTRKCP-274 describes an issue discovered when running with LRT for run-3 validations where the number of shared hits on tracks from the primary pass were increase when enabling LRT. Further investigation showed that this issue was also present in release 21 and impacts LRT as well as back-tracking.

This MR addresses part of this issue, where the use of the PRDAssociation was not propagated in the correct order when setting all the tools for the TrackFinder as the tools were initialised before setAssociation() was executed. The useAssociation flag in the SiTrajectoryElement_xk was thereby always false.

The fix seems to give the expected results, minor reduction to LRT efficiency but reduction of fakes by 50%. Further details are on: https://its.cern.ch/jira/browse/ATLIDTRKCP-274 showing small scaled validation tests. It would be great to get this merged and we can follow up on the minor that's remaining to understand.

FYI @sroe @goblirsc

InnerDetector/InDetRecTools/SiCombinatorialTrackFinderTool_xk/src/SiCombinatorialTrackFinder_xk.cxx

@@ -283,7 +283,15 @@ void InDet::SiCombinatorialTrackFinder_xk::newEvent(const EventContext& ctx, SiC
 void InDet::SiCombinatorialTrackFinder_xk::newEvent
 (const EventContext& ctx, SiCombinatorialTrackFinderData_xk& data, Trk::TrackInfo info, const TrackQualityCuts& Cuts) const
 {
-  if (not data.isInitialized()) initializeCombinatorialData(ctx, data);
+  
+  if (not data.isInitialized()) {
+    //Check if to use PRDAssociation before initializing all the tools
+    int useasso;
+    if(!Cuts.getIntCut   ("UseAssociationTool"  ,useasso      )) useasso = 0;
+    data.tools().setAssociation(useasso);
+
+    initializeCombinatorialData(ctx, data);
+  }
 
   newEvent(ctx, data);
   data.trackinfo() = info;