Various fixes for source-position information

[QuietMisdreavus]

This PR groups together a handful of fixes for source-position information. They fix a handful of related bugs where a node would have an invalid sourcepos range where the ending position turned out to be before the starting one. The issues being fixed include:

1. When an autolink (<http://example.com>) occurred in a block with an indentation offset, the source position range didn't account for the block offset, creating a range that was shifted over from the actual link range. This was fixed by properly applying the block offsets when creating source-position ranges for autolinks.
2. HTML comments occurring in their own block would fail to update the ending line in the parser, creating a negative-size range for the block. This was fixed by checking blocks to ensure that the parser's end-line matched the block's end line before finalizing the source position.
3. When a link node was given its source position, the parser assumed that it would never span multiple lines, creating a situation where the source-position range for a link that did span multiple lines would be erroneous or invalid. This was fixed by properly loading the parser's current line as the "ending line" for the link's source position.

[martincizek]

@QuietMisdreavus Can I ask you for checking if this is fixed in PR #223? I know it's a little weird that I came with a overlaping PR, but it has origins in an extensive work in upstream commonmark#298 and we've later added fixes for the gfm extensions too. So I think we should merge our effort. :)