This relies on an external file that apparently was generated many many years ago with a long-lost script. The data in there is outdated and not optimized for fast lookup.
This could be replaced by a new generator consuming CLDR data and outputting static data that can be compiled in and allows allocation-free in-place lookup.
Sune started to create a CLDR-based generator: https://cgit.kde.org/scratch/sune/countrieslistcldr.git/
After reading my kcontacts code, the file and the file generated from countrieslistcldr.git's generate app, besides the fact that the file is 10 times bigger because 10 times as much data, there might be an issue that in the "original" file, it seems to be english first, then other languages, for each country. I haven't reproduced that part of it.
For ISO to localized country name mapping, there's "iso-codes", providing .mo files for country names (among other stuff), trivial to integrate (i18nd("iso_3166-1", strbuf.leftRef(pos).toUtf8().constData());), and that's already used elsewhere. So we just need to look at an optimized way to map country names to ISO codes.