TrkEventTPCnv: Work around isSane() failue.

Just like for materialEffects, in some saved tracks the measurement
object also can trip isSane().  This happens, for example,
in the MuonCombineConfig unit test.  Work around by effectively bypassing
the isSane() test here in the same way we did for materialEffects.

Tracking/TrkEventCnv/TrkEventTPCnv/src/TrkTrack/TrackStateOnSurfaceCnv_p3.cxx

 /*
-  Copyright (C) 2002-2017 CERN for the benefit of the ATLAS collaboration
+  Copyright (C) 2002-2021 CERN for the benefit of the ATLAS collaboration
 */
 
 //-----------------------------------------------------------------------------
@@ -32,17 +32,19 @@ persToTrans( const Trk::TrackStateOnSurface_p3 *persObj, Trk::TrackStateOnSurfac
   const Trk::MaterialEffectsBase* materialEffects = createTransFromPStore( &matBaseCnv, persObj->m_materialEffects, log );
 
   // There were some tracks saved that violate the isSane test in
-  // TrackStateOnSurface.  If we were to pass materialEffects to this ctor
-  // then we would trip that assertion.  However, we want to preserve
-  // the previous behavior of the tp converters, which did allow reading
-  // such tracks.  So defer setting the material pointer until
-  // after the checks,
-  *transObj = Trk::TrackStateOnSurface (meas,
+  // TrackStateOnSurface.  If we were to pass meas or materialEffects
+  // to this ctor then we would trip that assertion.  However,
+  // we want to preserve the previous behavior of the TP converters,
+  // which did allow reading  such tracks.  So defer setting these pointers
+  // until after the checks,
+  *transObj = Trk::TrackStateOnSurface (nullptr,
                                         trackParameters,
                                         fitQoS,
                                         nullptr,
                                         persObj->m_typeFlags);
+  transObj->m_measurementOnTrack = meas;
   transObj->m_materialEffectsOnTrack = materialEffects;
+  transObj->setFlags();
 }