refactor: consolidate ~70 scattered constants into 6 new config files (phase 6)

Create 6 domain-focused config files in src/config/ to centralize operational
tuning knobs that were scattered across 15+ source files:

- server-timing.ts: SSE batching, state debounce, scheduled runs, error recovery
- auth-config.ts: session TTL, rate limits, hook timeout
- tunnel-config.ts: QR token rotation, tunnel process lifecycle
- terminal-limits.ts: input length, terminal cols/rows, session name
- ai-defaults.ts: AI model string, idle/plan context limits
- team-config.ts: poll interval, cache sizes

Eliminates all cross-file duplicates:
- AI model string: 5 occurrences → 1 (config only)
- STATS_COLLECTION_INTERVAL_MS: 2 → 1
- timeout: 10000 in hooks-config: 6 → 0 (now HOOK_TIMEOUT_MS)
- MAX_TRACKED_AGENTS shadow: removed, imports from map-limits

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

src/ai-plan-checker.ts

@@ -29,6 +29,7 @@ import {
   type AiCheckerResultBase,
   type AiCheckerStateBase,
 } from './ai-checker-base.js';
+import { AI_CHECK_MODEL, AI_PLAN_CHECK_MAX_CONTEXT } from './config/ai-defaults.js';
 
 // ========== Types ==========
 
@@ -44,8 +45,8 @@ export type AiPlanCheckState = AiCheckerStateBase<AiPlanCheckVerdict>;
 
 const DEFAULT_PLAN_CHECK_CONFIG: AiPlanCheckConfig = {
   enabled: true,
-  model: 'claude-opus-4-5-20251101',
-  maxContextChars: 8000,
+  model: AI_CHECK_MODEL,
+  maxContextChars: AI_PLAN_CHECK_MAX_CONTEXT,
   checkTimeoutMs: 60000,
   cooldownMs: 30000,
   errorCooldownMs: 30000,