Fix: deadlock in /clear endpoint — move write_status() outside events_lock

write_status() calls get_current_state() which re-acquires events_lock.
Calling it inside the `with events_lock:` block caused the same thread to
block forever (threading.Lock is not reentrant). Any /clear request would
hang the aggregator indefinitely, breaking detector recovery flow.

Bump to v2.3.2.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

index.html

@@ -216,7 +216,7 @@
 		<script>
 			const emoteEl = document.getElementById("emote");
 			const messageEl = document.getElementById("message");
-			const VERSION = "v2.3.1";
+			const VERSION = "v2.3.2";
 
 			// Sound system
 			let audioCtx = null;