extract Output trait: decouple AI core from Telegram

- Add src/output.rs with Output trait and 3 implementations:
  TelegramOutput (streaming via draft/edit), GiteaOutput (comments),
  BufferOutput (for worker/tests)
- Refactor run_openai_with_tools and execute_tool to use &mut dyn Output
- Remove run_claude_streaming, invoke_claude_streaming, run_openai_streaming
  (dead code — only OpenAI-compatible backend is used now)
- Remove BackendConfig::Claude code path from handler
- stream.rs: 790 → 150 lines

src/life.rs

@@ -4,6 +4,7 @@ use teloxide::prelude::*;
 use tracing::{error, info, warn};
 
 use crate::config::{BackendConfig, Config};
+use crate::output::TelegramOutput;
 use crate::state::AppState;
 use crate::stream::run_openai_with_tools;
 use crate::tools::compute_next_cron_fire;
@@ -62,12 +63,13 @@ pub async fn life_loop(bot: Bot, state: Arc<AppState>, config: Arc<Config>) {
             } = config.backend
             {
                 let sid = format!("life-{chat_id_raw}");
+                let mut tg_output = TelegramOutput::new(bot.clone(), chat_id, true);
 
                 let result = tokio::time::timeout(
                     std::time::Duration::from_secs(LIFE_LOOP_TIMEOUT_SECS),
                     run_openai_with_tools(
-                        endpoint, model, api_key, messages, &bot, chat_id, &state, &sid,
-                        &config, true,
+                        endpoint, model, api_key, messages, &mut tg_output, &state, &sid,
+                        &config, *chat_id_raw,
                     ),
                 )
                 .await;