Description
What is the expected enhancement?
This would be a breaking change to the interface, but token_swapper currently returns (NodeIndex, NodeIndex) as its swaps for a graph g: impl NodeIndexable where it's arranged that g.to_index(n.index()) is the way of converting a n: NodeIndex back to the input graph's NodeId.
Given that the mapping of the NodeIndex back to G::NodeId must be possible and known to the token_swapper, it would be more convenient to use if token_swapper returned its answer in this form.
(side note: there's an internal HashMap that tracks this mapping that's unnecessary, given how the inner StableGraph is constructed. (side side note: that mapping needn't be a HashMap and the graph needn't be a StableGraph with holes - could use a compact Graph and a [G::NodeId] as the map back))