fix(review): four findings from adversarial Codex review round 1

- twitter: align backend selection with the two-phase findings pattern —
  an unauthenticated twitter-cli (warn) no longer blocks a fully-working
  OpenCLI (ok) further down the candidate list
- watch: use _is_newer_version like check-update (the != comparison kept
  the downgrade prompt this branch fixed elsewhere)
- uninstall: third copy of the skill-dir removal also gets the symlink
  guard (full `agent-reach uninstall` path)
- doctor: a stale active_backend from a previous check on the singleton
  channel no longer leaks into an errored result

+4 regression tests (162 total).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

agent_reach/doctor.py

@@ -19,15 +19,18 @@ def check_all(config: Config) -> Dict[str, dict]:
     for ch in get_all_channels():
         try:
             status, message = ch.check(config)
+            active = getattr(ch, "active_backend", None)
         except Exception as e:  # noqa: BLE001 — doctor must survive any channel
-            status, message = "error", f"体检异常：{e}"
+            # Channels are registry singletons: a stale active_backend from a
+            # previous check must not leak into an errored result.
+            status, message, active = "error", f"体检异常：{e}", None
         results[ch.name] = {
             "status": status,
             "name": ch.description,
             "message": message,
             "tier": ch.tier,
             "backends": ch.backends,
-            "active_backend": getattr(ch, "active_backend", None),
+            "active_backend": active,
         }
     return results