AI: add enable_thinking reasoning toggle plumbed to llama.cpp

New optional SamplingOverride forwarded to llama-server as
chat_template_kwargs.enable_thinking (gates Qwen3-style reasoning
blocks). None leaves the template default; other backends ignore it.
Wired through the agentic-insight and chat-turn request bodies/handlers.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

src/ai/insight_generator.rs

@@ -3933,6 +3933,7 @@ Return ONLY the summary, nothing else."#,
             if let Some(ctx) = overrides.num_ctx {
                 c.set_num_ctx(Some(ctx));
             }
+            c.set_enable_thinking(overrides.enable_thinking);
             Box::new(c)
         } else {
             // Pure Ollama local.
@@ -4064,6 +4065,7 @@ Return ONLY the summary, nothing else."#,
         top_p: Option<f32>,
         top_k: Option<i32>,
         min_p: Option<f32>,
+        enable_thinking: Option<bool>,
         max_iterations: usize,
         backend: Option<String>,
         fewshot_examples: Vec<Vec<ChatMessage>>,
@@ -4091,6 +4093,7 @@ Return ONLY the summary, nothing else."#,
             top_p,
             top_k,
             min_p,
+            enable_thinking,
         };
         let backend = self.resolve_backend(kind, &overrides).await?;
         span.set_attribute(KeyValue::new("model", backend.model().to_string()));