refactor: channel router + scheduler use SDK clients

ChannelRouter: replace raw httpx with AsyncTurnstoneServer (single-node)
and AsyncTurnstoneConsole route methods (multi-node). Remove _post()
helper, _route_path(), and manual JSON construction.

Scheduler: replace raw httpx.Client with TurnstoneServer (sync). Lazy
per-node client cache with token rotation and stale client pruning.

Clean remaining Redis/MQ references from tests, docs, and config:
- test_tls_admin: redis.internal -> app.internal
- test_config: [redis] test data -> [database]
- docs/channels.md, console.md: rewrite for HTTP architecture
- docs/api-reference.md, openshell.md: remove stale diagram/Redis refs
- turnstone.example.toml: remove [redis] section
- .pre-commit-config.yaml: remove types-redis dependency
- QUICKSTART.md: remove bridge/Redis from deployment descriptions

docs/api-reference.md

@@ -2,8 +2,6 @@
 
 ## Overview
 
-> See also: [MQ Protocol diagram](diagrams/png/06-mq-protocol.png) | [Message Routing diagram](diagrams/png/07-message-routing.png) | [Redis Key Schema diagram](diagrams/png/08-redis-key-schema.png)
-
 `turnstone-server` exposes a browser-based chat UI backed by a
 **Starlette** ASGI application served by **uvicorn**. The server uses
 **Server-Sent Events (SSE)** via `sse-starlette` for real-time streaming
@@ -523,7 +521,7 @@ inactivity.
 
 Each SSE connection to a workstream receives its own delivery queue.  Events
 produced by the worker thread are fanned out to all registered listener queues,
-so multiple consumers (browser, bridge, console proxy, SDK) can connect
+so multiple consumers (browser, console proxy, SDK) can connect
 simultaneously and each receives every event.  On reconnect the client receives
 a full history replay, so no catch-up mechanism is needed.