fix: structured reasoning no longer breaks chat consumers

Normalize incoming reasoning at the shared heading boundary and completed
extraction, and flatten auxiliary content and reasoning before accumulation.
Reuse the existing text flattener with no implicit fragment separators.

Combine the earliest related work from zsuroy (#85791), the diagnosis and
patch from 2025hcsmile2010-hue (#104711, #104848), and completed extraction
work from liuhao1024 (#104717) as a slim redo, not a verbatim cherry-pick.

Two invariant tests exercise the real SDK and local HTTP fixture across
main streaming, Relay collection, auxiliary sync/async and completed output.
The standalone matrix improves from 32/84 to 84/84, preserving answers.

Co-authored-by: suroy <suroy@qq.com>
Co-authored-by: 2025hcsmile2010-hue <2025hcsmile2010@gmail.com>
Co-authored-by: liuhao1024 <sunsky.lau@gmail.com>

agent/auxiliary_client.py

@@ -6422,12 +6422,15 @@ class _ChatStreamAccumulator:
         if delta is None:
             return
         made_progress = False
-        piece = getattr(delta, "content", None)
+        from agent.message_content import flatten_message_text
+
+        piece = flatten_message_text(getattr(delta, "content", None), sep="")
         if piece:
             self.content_parts.append(piece)
             made_progress = True
         reasoning_piece = getattr(delta, "reasoning", None) or getattr(delta, "reasoning_content", None)
-        if reasoning_piece and isinstance(reasoning_piece, str):
+        reasoning_piece = flatten_message_text(reasoning_piece, sep="")
+        if reasoning_piece:
             self.reasoning_parts.append(reasoning_piece)
             made_progress = True
         # Evaluate both unconditionally: they accumulate state, not just progress.