fix: prevent extra <turn|> when assistant has content + tool_calls + continuation

The continues_into_next condition blocked continuation when the
assistant message had tool_calls (via 'not message.get(tool_calls)').
When tool responses were already rendered inline (ns_tr_out.flag),
the tool exchange is complete and the turn should continue — but
the condition forced a <turn|> close, producing an extra turn-end
marker between the first assistant's content and the continuation.

Fix: change the condition to allow continuation when tool responses
are present: (not tool_calls or ns_tr_out.flag).

Reported in https://github.com/vllm-project/vllm/pull/42776

chat_template.jinja

@@ -356,8 +356,7 @@
         {%- set continues_into_next = (
             role == 'model'
             and next_nt.role == 'assistant'
-            and not message.get('tool_calls')
-            and not ns_tr_out.flag
+            and (not message.get('tool_calls') or ns_tr_out.flag)
         ) -%}
 
         {%- if ns.prev_message_type == 'tool_call' and not ns_tr_out.flag -%}