specdelta: -dir names a set, so the move gate has an exit that is not a derogation - #875

`-dir` took one directory. The move gate refuses a run whose surface reaches
outside it and advised "Widen -dir" -- advice that could not be followed, for
the move this repository is actually making:

    OPENED  language/variables_spec.rb:366 ...
    OPENED  core/binding/implicit_parameters_spec.rb:3 ...

The 3.4.1 -> 4.0.9 surface spans core/ AND language/, and the only widening
available from `core` was the whole corpus -- library/, optional/,
command_line/, security/ and hours of them. So the only reachable exit was
-partial, and every real answer about the move would have come out stamped
PARTIAL. A derogation that is the normal road is not a derogation, and a gate
whose honest exit nobody can reach is not a gate.

-dir now names a SET, comma-separated, and the refusal names the exact -dir
that would cover the surface instead of saying "widen":

    specdelta: the move 3.4.1 -> 4.0.9 touches 1 file(s) that -dir core does not measure:
        language/variables_spec.rb
      A number from this run would not be a number about the move. Either
        -dir core,language
      which measures the whole surface, or -partial to say so on purpose --
      the report is then stamped PARTIAL.

Top-level suites and no deeper: 50 exact directories would be unreadable and
would go stale with the corpus, while `core,language` is both the honest
surface and something a reader can retype. The suggestion is computed from the
census, so it reaches core/ and language/ and NOT library/.

Two things the set makes possible to get wrong, both tested:

  - a file reached through two named directories (`-dir core,core/set`) is
    measured ONCE; counted twice it is counted on both sides, and the totals
    then move with how the caller spelled -dir rather than with the engines;
  - `-dir core,` does not quietly become the whole corpus.

Refs #799

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>