feat(moderation): two-tier batch+individual fallback pipeline

- After a batch LLM call, any result flagged analysis_incomplete is
  immediately fanned out to an individual per-message fallback queue
- Batch hard-fail (result.ok=false) and unhandled exceptions now also
  route all affected messages to the individual queue instead of waiting
  behind the conversation error cooldown
- Individual queue runs fully parallel (fire-and-forget per message),
  de-duplicated by a Set<messageId> so no double-processing
- processIndividualFallback runs in the main process (no worker pool IPC
  overhead for a single-item call), with retryWithBackoff 2x/2-15s
- AnalysisQueueStatus gains activeIndividualRequests +
  individualInFlightCount fields for dashboard observability

src/moderation/types.ts

@@ -137,5 +137,9 @@ export type ModerationWsEvent =
 export interface AnalysisQueueStatus {
   queuedConversations: number;
   activeRequests: number;
+  /** Number of single-message fallback calls currently awaiting the LLM. */
+  activeIndividualRequests: number;
+  /** Number of message IDs sitting in the dedup set (in-flight or about to start). */
+  individualInFlightCount: number;
   lastError: string | null;
 }