util: try to fix link regex

[jo3-l]

#1796 changed the link regex to catch URL-encoded links such as https://%20@yagpdb.xyz. Unfortunately, it caused a regression where text after the URL is matched. The new URL regex matches on the entirety of this text, including the second line:

https://yagpdb.xyz
and this is matched too!

We try to tune the regex so that it catches URL-encoded characters while avoiding the regression. The regex before #1796, after #1796, and after this PR is:

before #1796:    (?i)([a-z\d]+://)([\w_-]+(?:(?:\.[\w_-]+)+))([\w.,@?^=%&:/~+#-]*[\w@?^=%&/~+#-])
after #1796:     (?i)([a-z\d]+://)([\w\W_-]+(?:(?:\.[\w\W_-]+)+))([\w\W.,@?^=%&:/~+#-]*[\w\W@?^=%&/~+#-])
                                      ^^ added         ^^ added      ^^ added             ^^ added
this PR:         (?i)([a-z\d]+://)([\w-._~:/?#\[\]@!$&'()*+,;%=]+(?:(?:\.[\w_-]+)+))([\w.,@?^=%&:/~+#-]*[\w@?^=%&/~+#-])
                                      ^^^^^^^^^^^^^^^^^^^^^^^^^ changed   ^^ reverted ^^ reverted        ^^ reverted

Explicitly, the problem from #1796 is that \w and \W together match on every character. We revert back to the old regex, but expand the character set in the second capturing group per https://stackoverflow.com/a/1547940 so that it still matches https://%20@yagpdb.xyz.