[remark-ping] Improve parsing to support Unicode

[StaloneLab]

Rewrite almost entirely the parser in remark-ping to get better support for Unicode characters in non-fenced pings. For instance, to ping Moté, the most natural way would be @**Moté**, which didn't work before this PR.

The detection of a ping has been changed as follow:

• before, we used a RegEx, that matched only \w characters;
• now, the package contains a complete, RegEx-independant parser, that gets the end of a ping matching characters from Unicode categories White Space, Close Puncutation, Final Punctuation, Initial Punctuation, Open Punctuation and Other Punctuation.

These changes should also make it easier to switch the package to micromark soon, as it already consist of a state-machine.

[StaloneLab]

Of course CI cries because of Node 10... Time to remove it, farewell!