fix: defer streaming iteration linebreak to prevent blank line stacking

Follow-up to 669c60a6 (cherry-pick of PR #2187, fixes #2177).

The original fix emits a "\n\n" delta immediately after every
_execute_tool_calls() invocation. When the model runs multiple
consecutive tool iterations before producing text (common with
search → read → analyze flows), each iteration appends its own
paragraph break, resulting in 4-6+ blank lines before the actual
response.

Replace the immediate delta with a deferred flag
(_stream_needs_break). _fire_stream_delta() checks the flag and
prepends a single "\n\n" only when the first real text delta
arrives, so multiple back-to-back tool iterations still produce
exactly one paragraph break.

run_agent.py

@@ -670,6 +670,9 @@ class AIAgent:
         # Internal stream callback (set during streaming TTS).
         # Initialized here so _vprint can reference it before run_conversation.
         self._stream_callback = None
+        # Deferred paragraph break flag — set after tool iterations so a
+        # single "\n\n" is prepended to the next real text delta.
+        self._stream_needs_break = False
 
         # Optional current-turn user-message override used when the API-facing
         # user message intentionally differs from the persisted transcript
@@ -3471,6 +3474,13 @@ class AIAgent:
 
     def _fire_stream_delta(self, text: str) -> None:
         """Fire all registered stream delta callbacks (display + TTS)."""
+        # If a tool iteration set the break flag, prepend a single paragraph
+        # break before the first real text delta.  This prevents the original
+        # problem (text concatenation across tool boundaries) without stacking
+        # blank lines when multiple tool iterations run back-to-back.
+        if getattr(self, "_stream_needs_break", False) and text and text.strip():
+            self._stream_needs_break = False
+            text = "\n\n" + text
         for cb in (self.stream_delta_callback, self._stream_callback):
             if cb is not None:
                 try:
@@ -6756,8 +6766,13 @@ class AIAgent:
                     _msg_count_before_tools = len(messages)
                     self._execute_tool_calls(assistant_message, messages, effective_task_id, api_call_count)
 
-                    # Signal iteration boundary to stream consumers to prevent text concatenation
-                    self._fire_stream_delta("\n\n")
+                    # Signal that a paragraph break is needed before the next
+                    # streamed text.  We don't emit it immediately because
+                    # multiple consecutive tool iterations would stack up
+                    # redundant blank lines.  Instead, _fire_stream_delta()
+                    # will prepend a single "\n\n" the next time real text
+                    # arrives.
+                    self._stream_needs_break = True
 
                     # Refund the iteration if the ONLY tool(s) called were
                     # execute_code (programmatic tool calling).  These are