fix(engine): all 177 tests pass with Python orchestrator

- Increment step_count and track tokens in __emit_event__("step_completed")
  so thread bookkeeping matches the old Rust loop behavior
- Remove double-counting of tokens in bootstrap (orchestrator handles it)
- Match nudge text to existing TOOL_INTENT_NUDGE constant
- Fix FINAL result propagation (use stored final_result, not VM return)

Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>

crates/ironclaw_engine/orchestrator/default.py

@@ -181,8 +181,8 @@ def run_loop(context, goal, actions, state, config):
             if nudge_enabled and nudge_count < max_nudges and signals_tool_intent(text):
                 nudge_count += 1
                 __add_message__("user",
-                    "You described what you'd do but didn't write code. "
-                    "Please write a ```repl code block to execute your plan.")
+                    "You expressed intent to use a tool but didn't make an action call. "
+                    "Please go ahead and call the appropriate action.")
                 continue
 
             # Plain text response - done

crates/ironclaw_engine/src/executor/loop_engine.rs

@@ -282,8 +282,10 @@ impl ExecutionLoop {
         // Post-cleanup: persist final state
         match result {
             Ok(orch_result) => {
-                self.thread.total_tokens_used += orch_result.tokens_used.total();
-                self.thread.total_cost_usd += orch_result.tokens_used.cost_usd;
+                // Token tracking is handled by __emit_event__("step_completed")
+                // and __llm_complete__ within the orchestrator, so no need to
+                // add orch_result.tokens_used here (would double-count).
+                let _ = &orch_result.tokens_used; // acknowledge field
                 self.clear_runtime_checkpoint();
                 self.persist_runtime_state(None, &mut persisted_event_count)
                     .await?;

crates/ironclaw_engine/src/executor/orchestrator.rs

@@ -597,6 +597,10 @@ fn handle_emit_event(
         "step_completed" => {
             let input = extract_u64_kwarg(kwargs, "input_tokens").unwrap_or(0);
             let output = extract_u64_kwarg(kwargs, "output_tokens").unwrap_or(0);
+            // Increment step count (mirrors the old Rust loop's step_count += 1)
+            thread.step_count += 1;
+            // Track token usage
+            thread.total_tokens_used += input + output;
             EventKind::StepCompleted {
                 step_id: StepId::new(),
                 tokens: TokenUsage {