Implement bidi_should_transform()

This now returns true if any of the characters in the text belong to
the Arabic or Hebrew alphabet, or are one of the Unicode directional
formatting characters. This is just so the bidi machinery doesn't have
to run 100% of the time for 100% of the languages. I will also make it
so the Arabic language pack, as well as custom levels, have an RTL
attribute that always enables bidi (and does things like
right-alignment in textboxes and other design-flipping)

desktop_version/src/FontBidi.cpp

@@ -265,8 +265,39 @@ void bidi_destroy(void)
 
 bool bidi_should_transform(const char* text)
 {
-    // TODO
+    /* Just as an optimization, only run the whole bidi machinery if the
-    return true;
+     * language is actually an RTL one, _or_ if an RTL character is found. */
+
+    const char* text_ptr = text;
+    uint32_t ch;
+    while ((ch = UTF8_next(&text_ptr)))
+    {
+        // The standard Hebrew and Arabic blocks
+        if (ch >= 0x590 && ch <= 0x77F) return true;
+
+        // Extended Arabic B and A
+        if (ch >= 0x870 && ch <= 0x8FF) return true;
+
+        // LEFT-TO-RIGHT MARK and RIGHT-TO-LEFT MARK
+        if (ch == 0x200E || ch == 0x200F) return true;
+
+        // Some other directional formatting: LRE, RLE, PDF, RLO, LRO
+        if (ch >= 0x202A && ch <= 0x202E) return true;
+
+        // The more recent isolates: LRI, RLI, FSI, PDI
+        if (ch >= 0x2066 && ch <= 0x2069) return true;
+
+        // Hebrew presentation forms
+        if (ch >= 0xFB1D && ch <= 0xFB4F) return true;
+
+        // Arabic presentation forms A
+        if (ch >= 0xFB50 && ch <= 0xFDFF) return true;
+
+        // Arabic presentation forms B
+        if (ch >= 0xFE70 && ch <= 0xFEFE) return true;
+    }
+
+    return false;
 }
 
 const char* bidi_transform(const char* text)