fix: increase max_tokens to 8192 for R1 reasoning overhead

R1 models use 500-2000 tokens for <think> blocks before the final
response. 4096 was too tight — the model would exhaust the budget
mid-thought and never emit the JSON.

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

src/claude.rs

@@ -63,7 +63,7 @@ impl ClaudeClient {
     ) -> Result<(String, Option<Usage>)> {
         let body = MessagesRequest {
             model: self.model.clone(),
-            max_tokens: 4096,
+            max_tokens: 8192,
             system: system.to_string(),
             messages: messages.to_vec(),
         };