feat(discord-bridge): comenzi slash in loc de prefixul !

Comenzile devin application commands inregistrate pe guild (sync instantaneu,
spre deosebire de cel global care dureaza ~1h): /new [fork], /cd <cale>,
/model <sonnet|opus> cu Choice, /status, /stop, /cleanup [force], /help.

- allowlist-ul se aplica identic la interactiuni (check_ids comun, ca sa nu
  existe a doua implementare care diverge); refuz efemer, fara executie
- fiecare comanda face defer() inainte de lucru — altfel Discord marcheaza
  interactiunea esuata dupa 3s desi comanda a rulat
- sync tolerant: la esec (lipsa scope applications.commands) botul porneste
  normal si logheaza linkul de reinvitare necesar
- mesajele obisnuite raman neschimbate, inclusiv steering-ul mid-tur
- linkul de invitatie primeste scope=bot%20applications.commands; referintele
  la ! din ops/ si documentatie trecute pe /

Verificat in productie: 7 comenzi inregistrate pe guild, citite inapoi din API.
Suita: 296 passed cu discord.py, 293 passed + 3 skipped fara.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B29CApsP1JkSdjYaGaHpE7

proxmox/lxc171-claude-agent/discord-bridge/ops/claude-discord.service

@@ -37,7 +37,7 @@ ExecStart=%h/.claude-discord/venv/bin/python /workspace/romfastsql/proxmox/lxc17
 # ESENTIAL: la stop/restart omoara TOT cgroup-ul, nu doar procesul principal.
 # Fara asta procesele `claude` (406 MB RSS masurat) raman in urma si umplu
 # containerul. Nu prinde totusi ce s-a desprins din cgroup — pentru aia exista
-# comanda `!cleanup` (cleanup.py).
+# comanda `/cleanup` (cleanup.py).
 KillMode=control-group
 KillSignal=SIGTERM
 TimeoutStopSec=30