Fixed regression in processing IfcEdgeCurve with same EdgeStart and EdgeEnd

[ziyike]

see https://standards.buildingsmart.org/IFC/RELEASE/IFC4/ADD2_TC1/HTML/schema/ifctopologyresource/lexical/ifcedge.htm in "Attribute Definitions", for param 2 "EdgeEnd": The same vertex can be used for both EdgeStart and EdgeEnd. This appears to be a regression comparing to older versions of this function (InitAdvancedFaces) which uses XbimEdge to do the trimming and handled this situation.