feat(pm): report a [Decision]-titled card sitting in the dispatch pool (H54) (#17197)

`pm:queue` is the dispatch pool: a PM seat reads it as capacity and picks
from it. A card whose title opens `[Decision]` declares its deliverable is a
ruling, and a PM seat may not rule a contract or product decision — so the
one act the pool invites is the one act forbidden on that card. The filing
card measured the cost on one lane twice: 8 read as dispatchable and 1 was,
then 21 read and at most 3 were.

H54 is a report-only row over the `pm:queue` label page this sweep's loop
already consumes: one label and one title, both fields of a payload already
paid for, so the row costs no request and its population is covered by
construction. It fires on an OPEN issue (never a PR, never a closed card)
carrying `pm:queue` whose title, after optional leading whitespace, begins
with `[Decision]` — anchored at the start, never a substring.

The row reports the DISAGREEMENT between two carriers and hands it to the
triage seat, which owns the second one. It never relabels, never treats the
prefix as the state (the prefix is the filer's, the state is triage's), and
needs no new label: the vocabulary already carries the state.

H26's deferred-target note said this shape was deliberately not reported
because a title heuristic would make the sweeper guess at intent. That stance
was about H26 inferring a block's fate from its target's title and it stands
— H26 is not widened. Both prose sites now point at H54 so the file does not
contradict itself; two self-test cases pin the pointers and pin that H26's
own criterion survives at both.

Self-tests: 43 cases — a positive control plus six negatives that each differ
from it in exactly one feature, driven over a length-pinned table so a
fixture that stops being exercised fails.

Claude-Session: https://claude.ai/code/session_01MoTv7pn338AZ71owsp19gQ

Co-authored-by: Claude <noreply@anthropic.com>

Author: os-justin