fix(ci): parse gh api --paginate output as NDJSON in close_duplicate_issues

The old fetch_open_issues assumed gh api --paginate emitted one JSON value
per line and split on newlines before calling json.loads. In reality gh
--paginate concatenates the raw JSON array response bodies (with embedded
whitespace/newlines), so any body large enough to contain a real newline
inside a string field left json.loads to parse a truncated slice, blowing
up with 'Unterminated string starting at ...'. Every issues-triggered
Check Duplicate Issues run has been failing because of this.

Fix by asking gh to unwind each page with --jq '.[]', which emits exactly
one JSON object per line and paginates safely, then json.loads each line.

Co-authored-by: Krrish Dholakia <krrish-berri-2@users.noreply.github.com>

.github/scripts/close_duplicate_issues.py

@@ -47,22 +47,11 @@ def fetch_open_issues(repo: str | None) -> list[dict]:
         )
     else:
         endpoint = "repos/{owner}/{repo}/issues?state=open&per_page=100&sort=created&direction=asc"
-    cmd = ["api", "--paginate", endpoint]
 
-    raw = gh(*cmd)
-    # gh --paginate concatenates JSON arrays, so we may get multiple arrays
-    issues = []
-    for line in raw.strip().splitlines():
-        line = line.strip()
-        if not line:
-            continue
-        parsed = json.loads(line)
-        if isinstance(parsed, list):
-            issues.extend(parsed)
-        else:
-            issues.append(parsed)
+    raw = gh("api", "--paginate", endpoint, "--jq", ".[]")
+
+    issues = [json.loads(line) for line in raw.splitlines() if line.strip()]
 
-    # Filter out pull requests (they also appear in the issues endpoint)
     return [i for i in issues if "pull_request" not in i]