As part of the "improve SBE performance" project, we have been working to eliminate the remaining uses of the TraverseStage in the code base, instead favoring the traverseF and traverseP builtin functions for doing array traversal. The reason is that we have found that plans constructed to implement MQL queries using traverseF/traverseP tend to have substantially better performance than those constructed using TraverseStage. Note that the Bonsai optimizer will also use traverseF/traverseP instead of TraverseStage when doing ABT lowering.

Once TraverseStage is no longer used, there is probably not a huge motivation to keep it around in the code base. We should consider deleting it.