• rework AddPhiArgsToSuccessors to not redundantly request PhiArgs
  • either just use regular (non-eh aware) enumerator up top, or revise to not check for try entry successors below
• change logic in AddPhiArg to assert (in debug) if a redundant add request is made

If that change does not regain the TP lost in #85546, then

• revise GT_PHI_ARG so that it can optionally hold a block or a block list
• revise AddPhiArg to do proper block maintenance, and RBO/RangeCheck to handle the list case
• consider using a side table (at least during SSA build) to avoid having to search for PHIs/PHI Args