The tokenizer produces an end-tag token from the bytes below. How does the tree builder handle this </br> in the 'in body' insertion mode?