Have wcwidth() return 0 for marks instead of -1

Since kitty always treats marks as combinig chars, this allows us to
remove a few unnecessary branches

gen-wcwidth.py

@@ -217,9 +217,8 @@ def gen_wcwidth():
         p('\tswitch(code) {')
 
         non_printing = class_maps['Cc'] | class_maps['Cf'] | class_maps['Cs']
-        add(p, 'Null', {0}, 0)
+        add(p, 'Marks', marks | {0}, 0)
         add(p, 'Non-printing characters', non_printing, -1)
-        add(p, 'Marks', marks, -1)
         add(p, 'Private use', class_maps['Co'], -3)
         add(p, 'Text Presentation', emoji_categories['Emoji'] - emoji_categories['Emoji_Presentation'], 1)
         add(p, 'East Asian ambiguous width', ambiguous, -2)

kitty/emoji.h

@@ -1,4 +1,4 @@
-// unicode data, built from the unicode standard on: 2018-02-04
+// unicode data, built from the unicode standard on: 2018-02-05
 // see gen-wcwidth.py
 #pragma once
 #include "data-types.h"

kitty/screen.c

@@ -278,7 +278,7 @@ unsigned int
 safe_wcwidth(uint32_t ch) {
     int ans = wcwidth_std(ch);
     if (ans < 0) ans = 1;
-    return MIN(2, ans);
+    return ans;
 }
 
 static inline void
@@ -1445,8 +1445,7 @@ screen_wcswidth(Screen UNUSED *self, PyObject *str) {
     unsigned long ans = 0;
     for (i = 0; i < len; i++) {
         char_type ch = PyUnicode_READ(kind, data, i);
-        bool is_cc = is_combining_char(ch);
-        ans += is_cc ? 0 : safe_wcwidth(ch);
+        ans += safe_wcwidth(ch);
     }
     return PyLong_FromUnsignedLong(ans);
 }

kitty/unicode-data.c

@@ -1,4 +1,4 @@
-// unicode data, built from the unicode standard on: 2018-02-04
+// unicode data, built from the unicode standard on: 2018-02-05
 // see gen-wcwidth.py
 #include "data-types.h"