chore: sharpen syft-pr-review on proportionality, self-contained bullets and duplication

[koenvanderveen]

Summary

Five pieces of feedback on the review documents the skill produces. Three were rules the skill already had and did not enforce hard enough, so those are sharpened in place rather than joined by a near-duplicate neighbour.

Changes

Sharpened rules

• Length follows the code → Length and position follow the code — makes explicit that word count and placement are both claims about how much something matters, and names the case that prompted it: mechanical cleanups (a moved import, a renamed local, a deleted comment) get one short note at the end of their theme, never a bullet of their own.
• Every bullet stands alone — now says a theme heading is a filing label, not context. B — Errors that name both sides tells a reader nothing about what raises the error or who the two sides are, so every bullet under it names its own subject.
• One test, one bullet splits in two. The first keeps the shape; the new A test bullet carries its own context requires the condition, what runs, and what is then true, with the before/after wording: if a job list holds no unique name, the hint under the table renders a position instead of a name, not with no unique name left, the hint gives a position.

New code standards

• Two functions or methods doing substantially the same work — one belongs, called from both places.
• A test name that pads one that works: length is free when every word earns it, so drop articles and filler. test_no_hint_when_do_owns_no_jobs, not test_no_hint_when_the_do_owns_none_of_the_jobs.

Removed

• The semicolon clause in the test rule, which Scannable, not prose already covers.

Testing

193 lines / 1988 words / 19 rules, inside the 220 / 2200 / 20 budget. pre-commit passes. Regenerated the review on #9497 with it — see the comment there.