Add attention_bias argument in transformer block and transformer layer modules, addressing change in MCore

Add attention_bias argument in transformer block and transformer layer modules, addressing change in MCore #6341