Emojis are used as avatars, they are not part of the posts
I would argue that emojis would fit within a technical definition of "text". They fit inside the text/* mimetype.
So do uuencoded and base64-encoded images.
...and we're right back where we started from.
True, but my terminal, web browser, phone, and text editor don't render base64-encoded images in-line, nor is base64 encoding part of the Unicode spec. Emojis behave much more like text than like images.