One thing here is that Unicode is normally processed as 16-bit integers. The 8051 isn't the perfect choice for this, since it can't handle a large case statement or array lookup into so large structures in a direct way.

But it is easy to transform between Unicode and UTF8, in which case you will get a multi-byte representation of the characters. Most of the latin characters will require one byte each. But it is also much easier for an 8-bit processor if potentially processing the font data as multiple character tables in the range 0..127 or 0..255. After preprocessing the received data, you can use code similar to what you already have for performing the character-to-bitmap conversion.

This will also make it easier to support subsets of Unicode and - when you get the time/need - later add support for cyrillic etc.

In the end, it's possible to see the problem as many dynamically switched "code pages" for the printer to support.