fix(convex): adicionar logs obrigatorios em cron jobs para evitar shape_inference errors

- Adicionar console.log no inicio de autoEndInactiveSessions (liveChat.ts)
- Adicionar console.log no inicio de cleanupStalePendingPolicies (usbPolicy.ts)
- Documentar problema de shape_inference e solucao em OPERATIONS.md (Secao 11)
- Atualizar .env.example com BETTER_AUTH_SECRET de 32+ caracteres

O shape_inference do Convex self-hosted falha ao unificar arrays vazios
(logLines: []) com arrays de strings (logLines: ["msg"]). Garantindo que
todo cron job produza ao menos um log, evitamos o conflito de tipos.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

convex/liveChat.ts

@@ -747,6 +747,8 @@ const INACTIVITY_TIMEOUT_MS = 5 * 60 * 1000
 export const autoEndInactiveSessions = mutation({
   args: {},
   handler: async (ctx) => {
+    // Log obrigatorio para evitar shape_inference errors com logLines vazios
+    console.log("cron: autoEndInactiveSessions iniciado")
     const now = Date.now()
     const cutoffTime = now - INACTIVITY_TIMEOUT_MS
 

convex/usbPolicy.ts

@@ -309,6 +309,8 @@ export const cleanupStalePendingPolicies = mutation({
     staleThresholdMs: v.optional(v.number()),
   },
   handler: async (ctx, args) => {
+    // Log obrigatorio para evitar shape_inference errors com logLines vazios
+    console.log("cron: cleanupStalePendingPolicies iniciado")
     const thresholdMs = args.staleThresholdMs ?? 3600000 // 1 hora por padrao
     const now = Date.now()
     const cutoff = now - thresholdMs