feat(seidroid-review): post each verdict as a new comment

[bdchatham]

What

Replaces the sticky upsert with a plain POST. Each review posts a new verdict
comment; earlier ones stay.

-if id="$(gh api ".../comments" --jq "map(select(.body | startswith(\"$MARKER\"))) | .[0].id // empty")"; then
-  if [ -n "$id" ]; then
-    gh api -X PATCH "repos/$REPO/issues/comments/$id" -f body="$body"
-  else
-    gh api -X POST  "repos/$REPO/issues/$PR/comments"   -f body="$body"
-  fi
-fi
+if gh api -X POST "repos/$REPO/issues/$PR/comments" -f body="$body" >/dev/null; then
+  posted=true
+fi

Why

Editing a comment in place leaves it at its original position in the
thread and notifies nobody. So a re-review did not move, did not notify, and
silently destroyed the previous verdict's text — three properties that
together make a verdict nearly indistinguishable from no verdict.

This is not theoretical. Diagnosing a broken reviewer today, on sei-load#96
and #90:

	your trigger	verdict body written	rendered at
#96	23:18:45	23:22:49	20:53
#90	23:18:46	23:24:06	21:19

Both verdicts were written 4–5 minutes after the trigger and rendered
eighteen hours up the page. Runs completed green, the bottom of the thread
showed nothing new, and the conclusion "seidroid is still broken" was wrong —
the review had worked and its output was invisible.

Why earlier verdicts stay

They are the record of what the review said before the author's fixes,
which is what a reader compares against. A long pull request accumulates a
few; that is the accepted cost, and it is the behaviour of the ai-review
tooling this replaced.

A hazard this removes

<!-- seidroid-review --> is also the marker on the unplaced-findings note
posted by the preceding job (line 1099). The old lookup took .[0] — the
oldest marker comment — so on a pull request where that note came first,
the verdict would have overwritten the note instead of the previous verdict.

Verification

actionlint reports the same 5 pre-existing findings as main
(lines 326/1001/1177); none from this change. YAML parses; the posted flag
and its unposted-verdict fallback path are untouched.

The consumer pin bump is
sei-load#97.

🤖 Generated with Claude Code

[cursor]

PR Summary

Low Risk
CI-only publishing behavior in the reusable workflow; no auth, secrets, or driver logic changes beyond how the verdict is posted.

Overview
seidroid-review no longer upserts a single sticky verdict comment on the pull request. Each successful review always POSTs a new issue comment, and older verdicts stay in the thread.

This fixes re-reviews that looked like they never ran: patching in place left the comment at its original position (often far up the thread), sent no notification, and replaced the prior verdict text. Comments in the workflow file document that trade-off and note that keeping history matches the old ai-review behavior.

The change also drops marker-based lookup (.[0] on <!-- seidroid-review -->), which could PATCH the unplaced-findings note instead of the last verdict because that marker is shared. Step naming and header comments are updated to describe posting a new comment rather than upserting one.

^{Reviewed by Cursor Bugbot for commit 0d59b8f. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

The swap from sticky upsert to a plain POST is correct and self-contained — set -e handling, the posted flag, and the unposted-verdict fallback are all preserved, and MARKER is still required elsewhere. The only leftovers are four surviving comments that still describe the verdict comment as upserted, one of which is a rationale that no longer holds.

Findings: 0 blocking | 2 non-blocking | 0 posted inline

Blockers

• None at the file/PR level.

Non-blocking

• .github/workflows/seidroid-review.yml:1039 — the "State the review's position" step justifies skipping an opinionless review with "a review cannot be edited later the way that comment is upserted, so an opinionless one is permanent clutter." After this PR the verdict comment is not upserted either, so the contrast that rationale rests on is gone: both artifacts are now permanent. The conclusion (don't post an opinionless review) may still be right, but the stated reason no longer is.
• Three other comments still say "upsert" for behaviour this PR removed: line 43 ("The posting step never upserts a placeholder when the driver produced no verdict"), line 1127 ("a no-verdict run never upserts a placeholder"), and line 1167 ("the upsert below is rejected" — it is now the POST that would be rejected). Given how much this file leans on its comments as the explanation of record, worth sweeping in the same change.