[fix] NFD normalization should be the default in normalize.c, not NFKD, as NFKD does some unwanted things like converting superscripts and the Latin-ASCII transliterator does a better, more thorough job while staying faithful to the original string

src/normalize.c

@@ -13,7 +13,7 @@ char *utf8_normalize_string(char *str, uint64_t options) {
 
     if (options & NORMALIZE_STRING_DECOMPOSE) {
         have_utf8proc_options = true;
-        utf8proc_options |= UTF8PROC_OPTIONS_NFKD;
+        utf8proc_options |= UTF8PROC_OPTIONS_NFD;
     }
 
     if (options & NORMALIZE_STRING_STRIP_ACCENTS) {