Avoid use of naughtyRetrieve in SCT_RodDecoder

• Avoid use of naughtyRetrieve in SCT_RodDecoder
  • Prepare a collection, fill the collection and add the collection to the container
• Use IDC_WriteHandle and addOrDelete instead
  • https://atlassoftwaredocs.web.cern.ch/guides/trigger/idc/#1-fill-via-idc_writehandle-lock
• Use SCT_RDO_Container instead of ISCT_RDO_Container
• Update SCT_RodDecoder::SharedData
  • Add std::unordered_map<IdentifierHash, std::unique_ptr<SCT_RDO_Collection>> rdoCollMap and `std::unordered_map<IdentifierHash, SCT_RDO_Container::IDC_WriteHandle> writeHandleMap' to keep SCT_RDO_Collection and IDC_WriteHandle.
  • Add setCollection method to centralize manipulations related to linkIDHash (=waferHash).
• Remove SCT_RodDecoderErrorsHelper& errorsCache from SCT_RodDecoder::makeRDO because it is now manipulated in SCT_RodDecoder::SharedData::setCollection

RunTier0Tests.py found some differences in HLT results. SCT deadsensors of four HLT tracks were unexpectedly reduced from 1 to 0. RNN score and its significance of an HLT tau jet was sligtly changed. RunTier0Tests.log diff-root-q221.ESD.log This is caused by SCT_FlaggedConditionTool. https://its.cern.ch/jira/browse/ATLASRECTS-5624 was created for this.

Tag @abarton