feat(llm): add Gemini CLI provider as primary; set qwen3.5:9b as default Ollama model

- Add GeminiCliProvider: shells out to `gemini -p` with --output-format json,
  injection-safe prompt passing, MCP server suppression via temp workdir,
  6-slot concurrency semaphore, 60s subprocess deadline
- Add --llm-provider, --llm-model, --llm-base-url CLI flags for per-call overrides
- Provider chain: Gemini CLI → OpenAI → Ollama → Anthropic
- Move LLM timing to dispatch layer (LLM: Xs on stderr)
- Default Ollama model: qwen3:8b → qwen3.5:9b (benchmark shows better schema extraction)
- Add noxa mcp subcommand
- Add docs/reports/llm-benchmark-2026-04-11.md (Gemini vs qwen3.5:4b vs qwen3.5:9b)
- Bump version 0.3.11 → 0.4.0

Co-authored-by: Claude <claude@anthropic.com>

config.example.json

@@ -1,12 +1,16 @@
 {
+  "$schema": "./config.schema.json",
   "_doc": [
     "Copy to config.json and remove fields you don't need.",
     "Secrets (api_key, proxy, webhook, llm_base_url) go in .env — NOT here.",
     "BOOL FLAG LIMITATION: once set to true here, cannot be overridden to false",
     "from the CLI for a single run (no --no-flag support). Use NOXA_CONFIG=/dev/null",
     "on the command line to bypass this config entirely.",
+    "LLM provider/model are optional overrides. Leave them unset to keep the",
+    "Gemini -> OpenAI -> Ollama -> Anthropic fallback chain intact.",
     "on_change is intentionally absent — it must remain a CLI-only flag.",
-    "Unknown fields are silently ignored, so this file works across noxa versions."
+    "Unknown fields are silently ignored, so this file works across noxa versions.",
+    "Set output_dir to write results to files instead of stdout."
   ],
 
   "format": "markdown",
@@ -27,8 +31,5 @@
   "path_prefix": null,
   "include_paths": [],
   "exclude_paths": ["/changelog/*", "/blog/*", "/releases/*"],
-  "use_sitemap": false,
-
-  "llm_provider": "gemini",
-  "llm_model": "gemini-2.5-pro"
+  "use_sitemap": false
 }