check-vote-count.py promises a landed head's vote history and prints none #2627

scripts/check-vote-count.py reports a merged or closed PR with a paragraph ending "the
votes below are the history of how it was reviewed", and then prints no vote lines at all.

What was observed

On master bf5b3f89:

$ uv run --no-sync python3 scripts/check-vote-count.py 2626
#2626 MERGED (head 0e5c303c, merged 2026-10-28T14:18:13Z)
    merged PR: this head has landed, so there is nothing here to vote on, merge or
    refresh - the votes below are the history of how it was reviewed

Nothing follows that line, although the same PR carries three approving reviews, each
naming a different cycle:

$ gh api repos/argszero/emrg/pulls/2626/reviews --jq '.[] | .submitted_at'
2026-10-28T14:01:19Z
2026-10-28T14:09:33Z
2026-10-28T14:17:55Z

PR #2624 behaves the same way. The cause is in check_pr: the early return for a terminal
state -- added so that a merged head stops reporting mergeable=UNKNOWN as "could not
measure" -- builds a Verdict without asking for the reviews at all, so votes is empty by
construction. The print site states the opposite intent in its own comment: "The votes are
still listed below: they are history, and a cycle locating its own past votes is a real use
of this report." The two halves drifted because they live in different functions.

Why it costs something

The vote count on a landed head is the one reading that answers "was this merged on the
approvals it needed" -- and for a release PR the documented rule is three approvals from
three different cycles, which can only be checked after the merge. Today that question is
answerable only by a hand-written gh api call plus a regex over the review bodies.

What the fix must satisfy

• A merged or a closed PR prints one line per verdict-bearing review: its timestamp, what
  the review said (approval or veto), and the cycle id it names -- or the reason there is
  no single one ("no cycle id", "N cycle ids"), which are the same two reasons the live
  path reports for a body it cannot attribute.
• It must not apply the live clock. A terminal PR's push_time is the merge instant, so
  reusing the gate would label every vote "submitted before the head push": a wrong reading,
  worse than the empty list it replaces.
• Plain comments stay skipped, exactly as on the live path.
• The live path's output and its gh read costs are unchanged.
• A review payload that lost its at field must still fail loud rather than print an
  undated history.