Always use unicode ellipsis when ellipsizing

We were already unconditionally using the unicode character when the
input string was not pure ASCII, leading to different behaviour in
depending on the input string.

systemd[1]: Starting printit.service.
python3[19962]: foooooooooooooooooooooooooooooooooooo…oooo
python3[19964]: fooąęoooooooooooooooooooooooooooooooo…oooo
python3[19966]: fooąęoooooooooooooooooooooooooooooooo…ąęąę
python3[19968]: fooąęoooooooooooooooooąęąęąęąęąęąęąęą…ąęąę
systemd[1]: Started printit.service.

src/basic/string-util.c

@@ -443,7 +443,7 @@ static char *ascii_ellipsize_mem(const char *s, size_t old_length, size_t new_le
         if (old_length <= 3 || old_length <= new_length)
                 return strndup(s, old_length);
 
-        r = new0(char, new_length+1);
+        r = new0(char, new_length+3);
         if (!r)
                 return NULL;
 
@@ -453,12 +453,12 @@ static char *ascii_ellipsize_mem(const char *s, size_t old_length, size_t new_le
                 x = new_length - 3;
 
         memcpy(r, s, x);
-        r[x] = '.';
-        r[x+1] = '.';
-        r[x+2] = '.';
+        r[x] = 0xe2; /* tri-dot ellipsis: … */
+        r[x+1] = 0x80;
+        r[x+2] = 0xa6;
         memcpy(r + x + 3,
-               s + old_length - (new_length - x - 3),
-               new_length - x - 3);
+               s + old_length - (new_length - x - 1),
+               new_length - x - 1);
 
         return r;
 }