SAFETY-44: keep attribution in-tree, add the third clause, and correct a claim I made for my own precondition

Review APPROVED f109153 and the split. Three asks, all applied, plus a
correction it caught in the round-4 commit message.

C STAYS IN-TREE. Attribution is a third property of the same comprehension -
`sorted()` decides which vanished key wins a destination, so a build renaming
`inverter-b` -> `garage` while DELETING `inverter-a` names the deleted key as
migrated and the one that really moved as VANISHED. A reader of the disclosure
paragraph wants it beside the other two rather than only in Trello. One
sentence, same block.

AND A FOURTH CLAUSE REVIEW MEASURED THIS ROUND: `and after.data.get(k) ==
value` survives the suite and is load-bearing. Without it the rule degenerates
to BARE-NAME matching - a deleted `unit/old-x/sign_reason` was reported as
having migrated into a `sign_reason` holding a COMPLETELY DIFFERENT SENTENCE,
and essentially every off-roster vanished `unit/*` key would be suppressed.
Added to the paragraph and to INSTR-01, so all of `moved_to` is covered by one
fixture design rather than two-thirds of one.

⚠️ AND THE POINT IS THAT THEY ARE FOUR CLAUSES OF **ONE COMPREHENSION**, WHICH
IS STRUCTURAL AND NOT A LIST OF CHORES. Four conditions, four different reasons
to exist, one fixture set that can reach only some at a time - so rebuilding a
fixture to reach clause N moves the others. That is exactly how round 4's
disarming happened, and it is what INSTR-01 has to design against rather than
patch around. Review called this the sharpest thing found in four rounds and it
is a fact about the function, not about the two clauses that exposed it.

⚠️ AND I OVERCLAIMED FOR MY OWN PRECONDITION. The round-4 message said the
strengthened M12 precondition was needed because "the exit code alone cannot
say the second thing". Review checked: on a clean run the finding regex matches
nothing, so the second conjunct is IMPLIED by the first today. It is
belt-and-braces, load-bearing only if the exit-0 requirement is ever relaxed.
WHAT ACTUALLY CLOSED THE FIRST DRAFT'S GAP IS `_m12_new` BEING A SET DIFFERENCE
- the mutation must ADD a finding - and that is the check the narrowing's
removal turns red. The wrong half was credited, and the comment now says which
half does the work.

Nit: the fixture's trailing newline and two backticks, lost last round.

`rehearse_roster.py` is still byte-identical to d1b7e3f - this round touches
only the test file's disclosure paragraph and the fixture's comment. 28/28 pass.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BwRLw7eKYj2byVdwB6kbjU

goodwe_controller/tools/rehearsal/state-both-latched.json

@@ -1,5 +1,5 @@
 {
- "_comment": "SAFETY-44 SYNTHETIC FIXTURE - not a copy of any live state file; this session cannot read that one and did not try. It exists for ONE control: a store in which TWO ROSTER UNITS ARE LATCHED AT THE SAME VALUE. state-legacy.json cannot serve it - it is pre-DEV-02 and flat, so after migration only the FIRST roster entry is ever latched, and a rule that suppresses a vanished key when another key holds the same value can never be exercised against it. That is exactly how M12's first draft came to pass on the UNMUTATED tree. Both units are latched by ONE site event, so the reasons match too, which is the arrangement in which the suppression is most dangerous.",
+ "_comment": "SAFETY-44 SYNTHETIC FIXTURE - not a copy of any live state file; this session cannot read that one and did not try. It exists for ONE control: a store in which TWO ROSTER UNITS ARE LATCHED AT THE SAME VALUE. `state-legacy.json` cannot serve it - it is pre-DEV-02 and flat, so after migration only the FIRST roster entry is ever latched, and a rule that suppresses a vanished key when another key holds the same value can never be exercised against it. That is exactly how M12's first draft came to pass on the UNMUTATED tree. Both units are latched by ONE site event, so the reasons match too, which is the arrangement in which the suppression is most dangerous.",
  "schema_version": 1,
  "maint_schema_version": 1,
  "auto": true,
@@ -7,4 +7,4 @@
  "unit/garage/sign_reason": "SIGN_DISAGREEMENT: grid meter reversed, both units",
  "unit/shed/sign_tripped": true,
  "unit/shed/sign_reason": "SIGN_DISAGREEMENT: grid meter reversed, both units"
-}
+}