fix(agent): detect repeated tool call loops and make max_iterations configurable

Local LLMs often get stuck calling the same tool repeatedly because they
lack the sophistication to synthesize results. The agent loop had no
safeguard — it re-executed whatever the model requested up to 10 times.

Add fingerprint-based loop detection: if the same tool+args combination
repeats 3 consecutive times, break the loop and return the last results.
Also add agents.max_iterations to the config schema so the iteration
limit is user-configurable (default: 10).

docs/plans/state.json

@@ -1027,7 +1027,7 @@
   },
 
   "overall_progress": {
-    "total_test_count": 1268,
+    "total_test_count": 1292,
     "all_tests_passing": true,
     "p0_completion": "3/3 (100%)",
     "p1_completion": "4/4 (100%)",
@@ -1047,4 +1047,4 @@
     "gmail_auth_cli": "flynn gmail-auth command implemented with OAuth2 flow, doctor check, config routed to Telegram",
     "next_up": "GSD Milestone: Operator DX — Phase 3 Plan 02 (Dashboard UI consuming metrics RPC). All phases P0-P8 and Tiers 1-4 complete. Setup wizard added. TUI fullscreen mode now has full tool access and proper display. Remaining gaps: Tier 4 channels (Signal, Matrix, Teams, Google Chat), Tier 5 deferred/niche items"
   }
-}
+}