fix(weixin): recover from ret=-2 with empty stale-context errors

[notwin]

Summary

• Treat ret=-2 / errcode=-2 with an empty errmsg as the stale-context variant already recognized for "unknown error".
• Remove the stale peer token from both memory and the disk-backed token store before the existing tokenless retry.
• Add regression coverage for helper classification, durable token deletion, and the full send retry path.

Why

Cron-initiated pushes to inactive Weixin DMs can return ret=-2, errcode=None, errmsg=None when the persisted context_token has gone stale. The current adapter misclassifies that response as a rate limit, opens the circuit breaker, and drops the push. A new inbound message refreshes the token and makes later sends work again.

This extends the existing stale-session fallback without changing explicit frequency/rate-limit handling.

Fixes #18100.

Tests

• pytest tests/gateway/test_weixin.py tests/cron/test_scheduler.py -q -o addopts= — 321 passed
• ruff check gateway/platforms/weixin.py tests/gateway/test_weixin.py
• git diff origin/main...HEAD --check

[notwin]

Closing this approach. We will not pursue a source-level change here.

[alt-glitch]

This was generated by AI during triage.

Duplicate of #35714: the earlier open fix is the canonical Weixin stale context_token recovery path.