Read Copilot's "Open Issue" Overview Block in pr_review.py #3170

The shape

pr_review.py wait exited 43 UNRECOGNIZED_REVIEWER_OUTPUT on a Copilot round whose ccr-overview-v2 body carries an "open issue" block:

summary: (N) open issue  (round <head>)

With markup stripped, the review body reads in this order:

<!-- ccr-overview-v2 -->
### 🟡 Changes recommended
Unresolved critical and moderate findings remain, including <topic> and <topic>.
<details open><summary>1 open issue</summary>
- [High severity icon] [<finding title>](<link to the inline thread>) · New
</details>
<details><summary>What changed in this PR</summary> ... file table ... </details>

The block's one entry links to an inline review thread that the round also opened. So unlike #3167's "previously missed" entries, this finding is already counted as a thread. Here the script read threads=1 unresolved=1 correctly, and the only unreadable part was the block itself.

The heading is singular ("1 open issue"). A plural form ("N open issues") is presumably the same block, and so is "No open issues", which #3167's body also shows.

Expected

Recognize the "open issue(s)" block as a list of the round's own open threads, each entry tied by its link to a thread the digest already counts. wait should then report the thread rather than stopping on an unrecognized shape. If an entry links to no thread on the head, it should be reported the way a suppressed finding is.

Seen on 2026-10-16 on a downstream private repository's develop -> main promotion pull request. The round had no "Review effort" line.

Related: #3167, the "previously missed issues" sibling block in the same overview.