Use wide character function for categorizing codepoints

Previously we used the is... rather than isw...
These functions were not supporting non ascii characters correctly

src/highlighters.cc

@@ -404,7 +404,7 @@ void expand_unprintable(const Context& context, DisplayBuffer& display_buffer)
                             end = buffer.iterator_at(atom_it->end()); it != end; ++it)
                 {
                     Codepoint cp = *it;
-                    if (cp != '\n' and iscntrl((int)cp))
+                    if (cp != '\n' and not iswprint(cp))
                     {
                         std::ostringstream oss;
                         oss << "U+" << std::hex << cp;

src/normal.cc

@@ -375,7 +375,7 @@ void view_commands(Context& context, int param)
 void replace_with_char(Context& context, int)
 {
     on_next_key_with_autoinfo(context, [](Key key, Context& context) {
-        if (not isprint(key.key))
+        if (not iswprint(key.key))
             return;
         ScopedEdition edition(context);
         Buffer& buffer = context.buffer();

src/unicode.hh

@@ -3,6 +3,7 @@
 
 #include <cstdint>
 #include <ctype.h>
+#include <wctype.h>
 
 namespace Kakoune
 {
@@ -29,13 +30,13 @@ enum WordType { Word, WORD };
 template<WordType word_type = Word>
 inline bool is_word(Codepoint c)
 {
-    return c == '_' or isalnum(c);
+    return c == '_' or iswalnum(c);
 }
 
 template<>
 inline bool is_word<WORD>(Codepoint c)
 {
-    return !is_blank(c) and !is_eol(c);
+    return not is_blank(c) and not is_eol(c);
 }
 
 inline bool is_punctuation(Codepoint c)