Possible optimizations

As udomdiff manages regardless the final swap through the "slow path",
which is irrelevant for a single swap operation, while the code size
related to this final step is quite heavy, I've decided to prefer a
reduced production size over optimizing this case which occurrs when:

  * there is only one operation to do, irrelevant real-world optimization
  * it's at the end of more complex operations, still irrelevant

Performance wise, it doesn't seem to be a crytical change, but it's easy
to also eventually rollback if nevessary, as the removed block has been
flagged within the current code.