Parse HTML recursively

[RobertDober]

This would superseed #356 and is inspired by #353.

Basic idea

let the scanner be a little be more intelligent and scan the following line

<div> hello<br /> <span lang="greek">αλφα</span></div>

as

   [
     OpenTag{div, []}, Text{hello}, VoidTag{br}, 
     OpenTag{span, [{lang, greek}]} Text{αλφα} CloseTag{span}
     CloseTag{div}]

[RobertDober]

Notes:

• Scanning will become slower, but parsing should become faster, given that scanning is done in parallel while parsing is not this should not have a negative impact of overall performance.

• Right now the scanner uses some rgxen that might be subject to DoS so this might be another advantage of rewriting the scanner here, as splitting into junks by ~r{<[^>]+>}, seems totally safe.

• Failing to close a tag somewhere will totally alter the rest of the output, however the error messages will indicate what to do.

[RobertDober]

Here as an example what to strive for: GFM output

which is

<span><em>a</em>hello
<hr /><strong>b</strong></span>

[RobertDober]

will be implemented with RobertDober/earmark_parser#7