Performance issue in the definition of answer_module, a09_DynamicMemoryNet/a8_dynamic_memory_network.py

Hello, I found a performance issue in the definition of answer_module, a09_DynamicMemoryNet/a8_dynamic_memory_network.py, tf.concat will created repeatedly(tensorflow 1.x) during program execution, resulting in reduced efficiency. I think it should be created before the loop in answer_module.

Looking forward to your reply. Btw, I am very glad to create a PR to fix it if you are too busy.