Fix tokenizer EOF error positions

[fb55]

I am trying to move parse5 to the upstream html5lib-tests repo (away from this fork). As a first PR to come from this effort, this PR corrects some tokenizer errors. The changes are in three categories:

1. Off-by-one errors for EOF errors. Most EOF errors already point at the column after the last character, with some exceptions. These exceptions were fixed.
2. Line breaks being ignored by some EOF errors. Similar to (1), these are the exception.
3. unknown-named-character-reference errors were missing entirely and have been added. Reverted.

[Ms2ger]

Sounds fine. I'll give it a few days for others to comment; ping me if I forget

[untitaker]

I believe absence of errors is correct because &not (without a semicolon) is a valid character reference. This is also in line with what Firefox does to the HTML document a&noti, which decodes as a¬i.

[untitaker]

In fact if you check the spec, &noti.. is the exact example they use to describe that edgecase: https://html.spec.whatwg.org/multipage/parsing.html#named-character-reference-state:

if the markup contains the string I'm &notit; I tell you in an attribute, no character reference is parsed and string remains intact (and there is no parse error).

[fb55]

Thanks a lot for flagging @untitaker. I've reverted the additions.

[untitaker]

error locations are not actually standardized, right? this is just to make the testsuite internally consistent?

[fb55]

error locations are not actually standardized, right? this is just to make the testsuite internally consistent?

That is correct.

[untitaker]

I am not a maintainer but this lgtm

[fb55]

@Ms2ger It would be great if you could have another look at this (as well as #145 if possible)!