Convert HTML attributes containing emoji unicode

[lentschi]

When I try to convert the following HTML with your lib's replace_unified method:

<a title="Hello 🧛‍♀️!">Hello</a> 👌 🧛‍♀️ 👌!

... it produces the following (invalid) HTML:

<a title="Hello <span class="emoji emoji-sizer" style="background-image:url(/emoji-data/img-apple-64/1f9db-200d-2640-fe0f.png)" data-codepoints="1f9db-200d-2640-fe0f"></span>!">
  Hello
</a>
<span class="emoji emoji-sizer" style="background-image:url(/emoji-data/img-apple-64/1f44c.png)" data-codepoints="1f44c"></span>
<span class="emoji emoji-sizer" style="background-image:url(/emoji-data/img-apple-64/1f9db-200d-2640-fe0f.png)" data-codepoints="1f9db-200d-2640-fe0f"></span>
<span class="emoji emoji-sizer" style="background-image:url(/emoji-data/img-apple-64/1f44c.png)" data-codepoints="1f44c"></span>!

(note the span in the title attribute)

Am I doing something wrong or does your library just not support HTML input? If the latter is the case, you could maybe mention in the README that it just supports text that has been html escaped prior to conversion or even do the escaping within the library. (Though that would be a breaking change.)

Btw. for me the expected behavior would be to leave the title attribute untouched as it cannot contain images.