Graphemes don't all have the same width, not even when you use a monospace font.
For latin characters it usually works find to assume the same width. But emojis,
japanese or chinese characters have have different width. There are even some
ultra wide characters like 𒐫 or ﷽. There is also a thing
called 'half-width' character. E.g. the japanese 'a' can be ア or ｱ.
Fixed by actually computing the width of graphemes and using pixel.
In stage 1 we use a glob pattern matching the file name.
Stage 2 (which I will maybe implement some day) might use some additional magic byte sequence for file type detection.
