Catch a hand-edited status even when the primitives also moved

[Chessing234]

scripts/validate.py explains that a stale status on a branch is harmless, but a contributor editing status directly is not, because regeneration silently reverts it. The check only fired when status moved and both informal_status and formal_status stayed put, so it misses the case where a PR edits a primitive and hand-edits status in the same change.

That is the likely shape of the mistake in practice: someone updates informal_status and helpfully "fixes up" status to match what they think it should be. If their value disagrees with the derived one, it passes validation and is thrown away on merge.

Reproduced against a one-problem base file (informal_status: open, status: open):

case	branch content	before	after
A	status hand-edited, primitives untouched	❌ error	❌ error
B	informal_status → disproved, status hand-edited to proved	✅ Validation OK	❌ error
C	informal_status → disproved, status left stale	notice, exit 0	notice, exit 0
D	informal_status → disproved, status correctly updated	silent, exit 0	silent, exit 0

Case B is the gap; C and D confirm the stale-but-untouched path and the correct path are unchanged, so this does not start rejecting ordinary edits.

By the time control reaches this branch status already disagrees with the derived value, so the only remaining question is whether the PR touched status at all. Dropping the two primitive-equality conditions is what expresses that.

Also checked for false positives on the real file: validate.py --base over the current data/problems.yaml (1217 problems) against main~5 and against main itself both report ✅ Validation OK.