fix: respect enabled=False in iterator protocols (__iter__ / __aiter__)

[gaoflow]

Summary

When Retrying(enabled=False) is used with the context-manager iterator pattern — for attempt in retrying: or async for attempt in retrying: — the enabled flag is silently ignored and the full retry machinery runs (multiple attempts, stop/wait evaluated). This is inconsistent with the wraps()-based path, which already handles enabled=False correctly by running the function exactly once and propagating any exception directly.

Reproducer:

import tenacity

call_count = 0
retrying = tenacity.Retrying(enabled=False, stop=tenacity.stop_after_attempt(5))
try:
    for attempt in retrying:
        with attempt:
            call_count += 1
            raise ValueError("fail")
except tenacity.RetryError:
    # WRONG: call_count == 5, and a RetryError is raised instead of ValueError
    print(f"call_count={call_count}")  # prints 5, should be 1

Fix

• BaseRetrying.__iter__: when not self.enabled, yield one AttemptManager, then re-raise the stored exception if the attempt failed — mirroring wraps().
• AsyncRetrying.__aiter__ / __anext__: same semantics via a one-shot done-flag.

Four regression tests are added (sync failure, sync success, async failure, async success).

This pull request was prepared with the assistance of AI, under my direction and review.

[mergify]

Queued — the merge queue status continues in this comment ↓.

[mergify]

Merge Queue Status

• ✅ Entered queue — 2026-06-25 07:39 UTC · Rule: default · triggered by rule autoqueue
• ✅ Checks skipped · PR is already up-to-date
• ✅ Merged — 2026-06-25 07:39 UTC · at d9da50047ea9df4a62da614f84cf033f46f31dd1 · squash

This pull request spent 8 seconds in the queue, including 1 second running CI.

Required conditions to merge