Add a dump-html-tokens utility that prints HTML tokenizer output in a
stable text format, along with focused tokenizer input and expectation
files.
Include silent and repeated-iteration modes so the utility can also be
used for local tokenizer throughput checks without changing the dumped
token format.
