convert_precision_to_size calculates how many bytes of a UTF-8 string we can print before we exceed a display width limit, but does it using a weird two-step:

1. Walk the string, calculating the number of code points we can print, and throw away the position we stopped at.
2. Walk the string again to convert the number of code points into a number of bytes (i.e. recalculate the position we stopped at).

Put like that, the fix is clear. Also, this is the only place code_point_index, the function that does step 2, is used, so we delete it.