fix(asyncpg): Use distinct span ops for cursor iteration and fetch to prevent N+1 false positives

[ericapisani]

When asyncpg iterates a server-side cursor (via CursorIterator.anext) or fetches rows in batches (via Cursor.fetch), each batch triggers BaseCursor._bind_exec or BaseCursor._exec, producing a burst of identical spans. Sentry's N+1 detector incorrectly flags these as N+1 queries.

Set sentry.op to db.cursor.iter for CursorIterator-driven spans and db.cursor.fetch for Cursor.fetch-driven spans so the backend can exclude them from N+1 detection.

Adds span_op_override_value parameter to record_sql_queries to propagate the override without changing existing callers.

This will require changes on the backend in order to not trigger n+1 alerts on these ops.

Fixes #6576
Fixes PY-2534

[linear-code]

PY-2534

[github-actions]

Codecov Results 📊

✅ 92048 passed | ⏭️ 6140 skipped | Total: 98188 | Pass Rate: 93.75% | Execution Time: 322m 24s

📊 Comparison with Base Branch

Metric	Change
Total Tests	📈 +52
Passed Tests	📈 +52
Failed Tests	—
Skipped Tests	—

All tests are passing successfully.

✅ Patch coverage is 93.33%. Project has 2414 uncovered lines.
✅ Project coverage is 89.85%. Comparing base (base) to head (head).

Files with missing lines (1)

File	Patch %	Lines
sentry_sdk/integrations/asyncpg.py	92.86%	⚠️ 2 Missing and 2 partials

Coverage diff

@@            Coverage Diff             @@
##          main       #PR       +/-##
==========================================
+ Coverage    89.85%    89.85%        —%
==========================================
  Files          192       192         —
  Lines        23745     23774       +29
  Branches      8198      8206        +8
==========================================
+ Hits         21334     21360       +26
- Misses        2411      2414        +3
- Partials      1344      1348        +4

---

Generated by Codecov Action

[wahajahmed010]

Review: #6609 — fix(asyncpg): distinct span ops for cursor iteration/fetch

Logic: ✅ Clean approach. Using ContextVar flags to distinguish cursor iteration vs fetch vs regular query execution, then overriding sentry.op accordingly. The N+1 detector can then filter out db.cursor.iter/db.cursor.fetch spans.

Tests: ✅ Comprehensive — tests both streaming and static span modes, both cursor iteration and fetch paths. Asserts exact span count (5 for prefetch=5, 2 for fetch(10)+fetch(10)).

Edge case: What if someone uses CursorIterator.__anext__ directly without going through async for? The monkey-patch on CursorIterator.__anext__ covers that. ✅

Concern: The ContextVar flags are module-level — if two cursors are active concurrently (e.g., in different tasks), the flags could cross-contaminate. However, ContextVar is task-local by design in Python, so this should be safe. ✅

Performance: Minimal overhead — just a ContextVar set/get per call. ✅

Verdict: Well-structured fix. Ship it.