fix(channels): raise doctor probe timeout 5s→15s (cold-TLS false ConnectTimeout)

[acedatacloud-dev]

What

channels doctor used a 5.0s httpx timeout for its authenticated probe
(GET /api/tasks/<probe>). A cold TLS handshake to a hosted WeChat gateway
(*.wisdom.acedata.cloud) takes ~6.6s on the first connect (measured), so
doctor reported [FAIL] unreachable: ConnectTimeout even though the gateway was
healthy and channels start connected fine. That false negative sends users
chasing a nonexistent connectivity problem during first-time setup.

Fix: raise the probe timeout 5.0 → 15.0s.

Evidence (live)

# before (5s): first-connect handshake overran the timeout
timeout=5.0  -> ConnectTimeout in 6.65s
timeout=10.0 -> 404 in 2.88s        # warm connection is fast

# after the patch, against the real gateway:
[ENABLED] wechat/wisdom-ebcdd3c54eb2: https://ebcdd3c54eb2.wisdom.acedata.cloud
  [OK] OK (token accepted, 43 bytes)

channels smoke and channels start were already succeeding against the same
gateway; only doctor mis-reported.

Scope

One-line timeout constant + comment. The existing doctor test monkeypatches
WeChatClient.__init__(…, *, timeout=…) and forwards the arg, so it is
unaffected (no test asserts the literal 5.0).

🤖 对抗评审

Reviewer: Codex (codex exec --sandbox read-only, cross-model). Both this
diff and the paired PlatformFrontend snippet fix were reviewed together for
correctness, wrong commands/env-var names, token leakage, and test breakage.

VERDICT: CLEAN — no findings.