ci(musa): use unified Qwen model paths

Author: hooray03

tests/models/qwen3_6/27b_tp4_eager_musa.yaml

@@ -1,7 +1,7 @@
-# Qwen3.6-27B Hygon smoke configuration (TP=2, eager).
+# Qwen3.6-27B MUSA smoke configuration (TP=4, eager).
 
 llm:
-  model: "/workspace/Qwen3.6-27B"
+  model: "/data/models/Qwen/Qwen3.6-27B"
   tensor_parallel_size: 4
   pipeline_parallel_size: 1
   max_model_len: 8192

tests/models/qwen3_6/35b_a3b_tp4_eager_musa.yaml

@@ -1,7 +1,7 @@
-# Qwen3.6-35B-A3B Hygon smoke configuration (TP=2, eager).
+# Qwen3.6-35B-A3B MUSA smoke configuration (TP=4, eager).
 
 llm:
-  model: "/workspace/Qwen3.6-35B-A3B"
+  model: "/data/models/Qwen/Qwen3.6-35B-A3B"
   tensor_parallel_size: 4
   pipeline_parallel_size: 1
   max_model_len: 8192