Add support for unicode email addresses (RFC 6530 and following)

[arnt]

This is as good as it can be using the builtin re module. Using the regex module would permit another improvement.

As it stands, this matches å as a combined character, which is by far the most common way to encode å. It does not match two things:

• 'a' + combining ring above
• ZWJ, which is required e.g. for the word 'sri' when written using sinhala, the most common script on Sri Lanka.

Switching to the regex module would permit detecting both of those, but they're rare cases and it's not obvious to me that catching some rare cases justifiies pulling in another module.

[tsutsu3]

I think it's a great idea.
However, since it behaves differently from linkify-it, I will not merge it.

[arnt]

I see, that makes sense to me. Please leave this open and I'll submit corresponding PRs to linkify-it and I suppose I might as well do linkify-it-rb too.