After thinking about it some, I don't like the way some of the solutions behave when you have \n at the end or beginning. To me, it would make more sense to ignore these. You can do that as a separate step, or integrate it in to the regex:

my $paragraphs = $comments_to_add =~
    s&(?=.*\S|\A)\s*(.*?)\s*?(?:\n(?!\r?\n)|\z)&<p>$1</p>&g;
[download]

This assumes you want <p></p> even around an empty comment. If that's not the case, it's easier:

my $paragraphs = $comments_to_add =~
    s&\s*(.+?)\s*?(?:\n(?!\r?\n)|\z)&<p>$1</p>&g;
[download]

I was dissatisfied with how complicated this turned out. If anyone has suggestions for making it simpler, I'd love to see.