Markdown lexer improvements

[Leistungsabfall]

This is a series of improvements to the Markdown Lexer:

• Fenced code was not recognized when it contained empty lines
• Inline code was not recognized after bulleted list entry (e.g. * `code` )
• Bold and italics were not recognized when surrounded with non-space char (e.g. (**foo**))
• There is no way to recognize code indented by 4 spaces using regex only - so remove the error-prone detection

[gerner]

I verified that the perf issue described in #1617 is fixed by this change with identical output from pygmentize on the test file. If this PR gets merged then my proposed fix in #1618 is unnecessary.