better translation workflow

[petitlapin]

As written in https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=432172, Li-ri uses a lot of space to store the translation (one image per text per language).

It would be better to have a standard gettext workflow and use SDL_ttf to display the text. A drawback would be we would use the same font (or almost) for all the texts (and we may lose all the actual translation as they are not kept apart, only in the images) but the advantages are way outstanding (standard flow, I would know how to add languages...).