fix(quotas): address review: resume claims, agent bucket rule, UI races

- A tool continuation refused for usage now releases the resume claim it
  took; before, retries got a 409 until the stale claim was reverted.
- Agent traffic is any row with an agent key or an agent id, so keyless
  agents and workflow nodes count toward the agent bucket, not direct.
- The user quota modal discards responses for a previously opened user.
- The usage meter shows every limited bucket, not only 'all'.
- Restore the class separator on the analytics stat card that a formatter
  run removed, and align the OpenRouter DeepSeek description with its rates.

docsgpt/api/v1/routes.py

@@ -258,6 +258,8 @@ def chat_completions():
 
     try:
         processor = StreamProcessor(internal_data, decoded_token)
+        # Set when this request took the resume claim, so a refusal can release it.
+        claimed_conversation_id = None
 
         if internal_data.get("tool_actions"):
             conversation_id = internal_data.get("conversation_id")
@@ -282,6 +284,7 @@ def chat_completions():
                     claimed_state=pending_state,
                 )
                 processor.conversation_id = conversation_id
+                claimed_conversation_id = conversation_id
             else:
                 # Compatibility fallback for old/completed conversations and
                 # clients that resend the full transcript without resumable
@@ -338,7 +341,10 @@ def chat_completions():
             )
 
         helper = _V1AnswerHelper()
-        usage_error = helper.check_usage(processor.agent_config, processor.decoded_token)
+        if claimed_conversation_id:
+            usage_error = helper.check_usage_on_resume(processor, claimed_conversation_id)
+        else:
+            usage_error = helper.check_usage(processor.agent_config, processor.decoded_token)
         if usage_error:
             return usage_error