Opening brackets incorrectly skipped when parsing (CommonMark ex. 551)

[pbodnar]

There is a negative side effect from the fix of #124: The CommonMark example 551 won't pass, because the opening brackets are skipped now:

> python -m test.specification
example:  523
markdown: '[\n ]\n\n[\n ]: /uri\n'
html:     '<p>[\n]</p>\n<p>[\n]: /uri</p>\n'
output:   '<p>]</p>\n<p>]: /uri</p>\n'

The example is mostly artificial and hopefully not used in practice, yet it would be good to have this fixed.

[pbodnar]

#130 was created in the meantime with more details, let's continue there...