Some UTF8 multibyte chars are splitted into invalid UTF8 sequences

[gfx]

p Greenmat::Markdown.new(
    Greenmat::Render::HTML.new,
    fenced_code_blocks: true
  ).render("```ム")

produces:

"<pre><code class=\"\xE3\x83\"></code></pre>\n"

which must be "<pre><code class=\"ム\"></code></pre>\n". Because \xE3\x83 is a invalid UTF8 sequence, the output result easily causes encoding errors (e.g. by Nokogiri).

Note: redcarpet also has the same bug.

[gfx]

There is a PR in redcarpet: vmg#464

[r7kamura]

@gfx We'll check it and move to solve this problem. Thank you for your reporting 👍