Headings inconsistent with the original Markdown renderer

[goblindegook]

I found a trio of issues with the rendering of headers where Parsedown is inconsistent with the original Markdown implementation:

• ####### not an h7 is rendered as a paragraph, while the original Markdown interprets it as <h6># not an h7</h6>.
• # # of levels (mind the space) is rendered as <h1>of levels</h1>. The original Markdown retains the extra hash as <h1># of levels</h1>.
• # # of levels # # is also rendered as <h1>of levels</h1>. The original Markdown retains the extra hashes both at the start and the end, <h1># of levels #</h1>.

These inconsistencies can be resolved by removing only the first n characters from the start of the string, n being the heading level. An additional trim(rtrim($line, '#')) should be performed --- the original Markdown implementation doesn't seem to care about the heading level when removing trailing hashes.

[erusev]

Thanks! The first is supposed to work like this. That's related to the CommonMark spec. The second and the third, however, aren't. I'll see what i can do.

[bkuhl]

#337 is also a PR related to streamlining headers and if both are implemented, would be paired well with a new release.