refactor(automod): remove regex classifier — LLM is the sole judge

Delete fastClassifier.ts (manual regex patterns for phone/email/IP/crypto/
spam/toxicity) and simpleFallback.ts. These hardcoded patterns were the
source of false positives (Discord emoji snowflakes matched phone_number,
URL digits matched phone, etc.) and produced heuristic verdicts whenever
the LLM failed.

New flow: Message → LLM (with conversation context, media evidence, user
reputation) → verdict. On LLM failure the message is marked 'error' and
retried by the recovery worker — no heuristic verdicts, ever.

Discord markdown tokens (custom emoji/mentions/timestamps) are normalized
to readable placeholders ([emoji:name], @user, @role, #channel, [time])
before reaching the LLM via discordTokens.ts.

services/discord-gateway/src/modules/ai-moderation/index.ts

@@ -1,9 +1,12 @@
 export { startPendingAIAnalysisWorker } from "./aiAnalyzer.js";
 export { runModerationAnalysis } from "./moderationOrchestrator.js";
 export { buildSystemPrompt } from "./moderationPrompt.js";
-export { runSimpleTextFallback } from "./simpleFallback.js";
+export { sanitizeDiscordTokens } from "./discordTokens.js";
 
-// ── New two-pass pipeline exports ──────────────────────────────────────────
-export { classifyMessage } from "./fastClassifier.js";
-export type { Layer1Result } from "./fastClassifier.js";
-export type { AnalysisInput, AnalysisResult, WorkerConfig, MessageBatch } from "./ai-analysis-worker.js";
+// ── Single-pass LLM pipeline exports ─────────────────────────────────────
+export type {
+  AnalysisInput,
+  AnalysisResult,
+  WorkerConfig,
+  MessageBatch,
+} from "./ai-analysis-worker.js";