add line tagging for each token/node

[syjer]

As this feature has an impact to the performances, it should be provided as an option+extended input processor.

[haumacher]

This is an essential feature, when using the parser as part of a HTML5 verifier/linter. Additionally, the parser should produce reasonable error messages/codes, if it detects a violation of the spec. Currently, only an empty parse error hook is called without context information.

I think, it is hard to provides this as performance-neutral optional feature, since it is a cross-cutting concern starting at tokenization and ending in the tree building phase.